The ingroup sample included multiple representative species of all five genera of the Pithecellobium clade, for a total of 20 of the 33 species (61%) of the group (Table 1). Species were sampled from across the geographical range of the Pithecellobium clade. The number of species sampled per genus relative to total diversity was as follows: Ebenopsis (2/3), Havardia (5/5), Painteria (3/3) Pithecellobium (8/18) and Sphinga (2/3). Although less than half of the species of Pithecellobium were sampled, these included species of the three most basal lineages of the genus, as identified in the morphological analysis of Barneby and Grimes (1997). We were unable to sample the endemic Antillean species of Pithecellobium (e.g. P. circinale (L.) Benth., P. cynodonticum Barneby & J.W. Grimes and P. histrix (A. Rich.) Benth.), as well as some species from Central America (e.g. P. furcatum Benth. and P. peckii Blake). Based on the results of previous phylogenetic studies, the outgroup was composed of representative species of nine more distantly related mimosoid genera (Table 1; Brown et al. 2008; Kyalangalilwa et al. 2013; Iganci et al. 2015; Koenen et al. 2020). Vachellia farnesiana (L.) Wight & Arn was used to root the analyses.

Thirty-two sequences were newly generated for this study, while 31 sequences were obtained from GenBank (www.ncbi.nlm.nih.gov/genbank); most of the latter were generated for the studies of Miller and Bayer (2001), Miller et al. (2003), Brown et al. (2008) and Duno de Stefano et al. (2021). The ETS dataset consisted of 34 sequences (21 new) from 33 species; the ITS dataset consisted of 29 (11 new) sequences from 30 species. GenBank accession numbers for all sequences are given in Table 1.

Total genomic DNA was extracted from fresh leaves collected from the living collection in the Roger Orellana Regional Botanical Garden of the Centro de Investigación Científica de Yucatán, A. C., from leaflet tissue collected in the field and dried with silica gel or from herbarium specimens deposited in the following Herbaria: CICY, FCME, MA, MEXU, MO, UCOL and ZEA (acronyms as in Thiers (2020[and onwards]). Total DNA (from fresh or herbarium material) was obtained with DNeasy Plant Mini Kits (QIAGEN Inc., Valencia, California), following the manufacturer’s specifications. Concentration and relative quality of DNA was evaluated using the protocol in Duno de Stefano et al. (2021).

Amplifications were performed in an Applied Biosytems Veriti 96 Well Thermal Cycler (Applied Biosystems, Foster City, USA). Volumes of reagents in PCR reactions (all reactions were brought to final volume by adding ultrapure water) and cycling conditions were as follows for the two DNA regions: 1) ETS: 30 μl of mix containing 3 μl 10X Buffer, 2.5 μl MgCl₂, 0.6 μl (~ 10 ng) each of primers, 4 μl Q solution, 1 μl 1.25 mM l-1 dNTP, 0.2 μl (1 U) TAQ polymerase, 2 μl (~ 10 ng) DNA; 94 °C for 3 min + 30 cycles (94 °C for 1 min + 60.5 °C for 1 min + 72 °C for 2 min) + 72 °C for 7 min; primers were 18S-IGS and 26S-IGS (Baldwin and Markos 1998) 2) ITS: 25 µl mix containing 2.5 µl 10X Buffer, 2.5 μl MgCl₂, 0.6 μl (~ 10 ng) each of primers, 4 μl Q solution, 1 μl 1.25 mM l-1 dNTP, 0.2 μl (1 U) TAQ polymerase, 2 μl (~ 10 ng) DNA; 94 °C for 3 min + 30 cycles (94 °C for 1 min + 60.5 °C for 1 min + 72 °C for 2 min) + 72 °C for 7 min; primers were Ac 12F and Ac 1290R from Miller and Bayer (2001). PCR products (and the primers used for amplifications) were sent to Macrogen Korea, Seoul, South Korea for sequencing.

Assembly and editing of sequences were carried out in BioEdit v.7.0.9 (Hall 1999). Each of the two partitions was aligned independently in the online version of MAFFT (Katoh et al. 2002, 2017; https://mafft.cbrc.jp/alignment/server/) using the default settings (gap opening penalty = 1.53 and offset value = 0.00). Following exploratory phylogenetic analyses employing maximum parsimony, the concatenated alignment matrix was analysed using Bayesian Inference as implemented in MrBayes v. 3.2.7 (Huelsenbeck and Ronquist 2001; Ronquist and Huelsenbeck 2003), with each partition (ETS and ITS) treated as independent and associated with its own nucleotide substitution model. The best fitting model for each partition was selected in jModelTest v. 2.1.7 (Darriba et al. 2012), based on the Akaike Information Criterion (AIC). The following substitution models were selected: ETS = TIM3+I+G and ITS = GTR+I+G. Bayesian analyses consisted of two independent Metropolis-coupled Markov Chain Monte Carlo (MCMC) runs, each starting from a randomly chosen tree and run for five million generations, with one tree sampled every 2000 generations. Twenty percent of the sampled trees were discarded as burn-in after evaluation of the output parameters generated by the Bayesian analysis in Tracer v.1.6 (Rambaut et al. 2014). The remaining sampled trees were summarised in a 50% majority-rule consensus tree, with clade posterior probabilities (PP, i.e. the proportion of trees containing particular clades) used to measure clade support.

Posterior Probabilities of < 0.95 were considered weakly supported, whereas PP of 0.95–1.0 were deemed to be well supported. The convergence of MCMC runs was assessed with Tracer v. 1.7.1 (Rambaut et al. 2014) verifying that the effective sample size (ESS) for all parameters was > 200 (Nascimento et al. 2017).

Our study shows both agreement and disagreement with well-supported results (PP ≥ 0.95 and/or likelihood or parsimony bootstrap ≥ 80%) of previous (Brown et al. 2008; Iganci et al. 2015; LPWG 2017; Koenen et al. 2020; Soares et al. 2021) and concurrent (Ringelberg et al. 2022) molecular phylogenetic studies. Our study is in agreement with all of these studies in strongly supporting the monophyly of the Pithecellobium clade. It also agrees with LPWG (2017) in supporting the monophyly of Sphinga and in the recovery of a clade grouping Havardia albicans, H. mexicana and H. pallens. With Ringelberg et al. (2022), which unlike the other studies sampled the key taxa Havardia campylacantha and Painteria elachistophylla, our study agrees in demonstrating the non-monophyly of both Havardia and Painteria, the latter due to the nesting of Ebenopsis within it and also agrees in supporting the monophyly of Pithecellobium.

Our results conflict with those of several studies by placing Pithecellobium as the sister taxon to Sphinga (Fig. 3). For example, the study of Iganci et al. (2015), although it sampled only single species of each genus (and none of Painteria), placed Havardia pallens sister to Pithecellobium dulce (PP = 0.99), to the exclusion of Sphinga acatlensis. The same result, but even more strongly supported (100% likelihood bootstrap), was obtained by Koenen et al. (2020), while Ringelberg et al. (2022) had H. pallens sister to Pithecellobium, but with more species sampled for the later genus. Interestingly, Koenen et al. (2020) recovered a clade within which Ebenopsis confinis was sister to H. pallens plus P. dulce (likelihood = 0.93, PP = 0.81), a result not recovered by any other analysis. The resolution of relationships within Pithecellobium is also somewhat conflicting between our study and those of LPWG (2017) and Ringelberg et al. (2022), although the overlap in sampling of species of the genus was limited between studies. For example, LPWG (2017) placed P. keyense in an unresolved position outside a clade comprising the other sampled species of the genus and Ringelberg et al. (2022) placed it sister to that clade, while in our phylogeny, P. keyense occupies a more nested position within Pithecellobium.

The causes of such conflict are unknown. Unlinked DNA regions used in the different studies may reflect different evolutionary histories, each with the potential to be differently impacted by evolutionary phenomena that cause phylogenetic conflict, such as gene duplication, hybridization and incomplete lineage sorting (Schrempf and Szöllősi 2020). Conversely, conflict may be attributable to statistical error, resulting from large differences in taxonomic sampling, the choice of sequence alignment criteria and/or the phylogenetic methods used, some of which may be more or less prone to phenomena, such as long branch attraction (Bergsten 2005). With respect to the present comparisons, we suspect that the latter is more prevalent since the other molecular phylogenetic studies of the Pithecellobium clade have less taxonomic sampling than our study. Moreover, we surmise that, in cases of conflict, our results are more compatible with previous taxonomic hypotheses, based on morphology.

Indeed, our study exhibits considerable agreement with the phylogenetic analysis of morphological data in Barneby and Grimes (1996, 1997), especially if the comparison is restricted to the clades from the morphological analyses that were supported by three or more putative morphological synapomorphies. For example, the monophyly of the alliance and the genera Ebenopsis, Pithecellobium and Sphinga, which were each strongly supported in our study, were also each supported by four to seven morphological synapomorphies in the analysis of Barneby and Grimes (1996, their fig. 12). Conversely, the two genera that were resolved as non-monophyletic in our phylogeny in conflict with Barneby and Grimes (1996), Painteria and Havardia, were respectively supported by only one and two putative morphological synapomorphies in the latter study. Within Pithecellobium, internal nodes supported by one or two morphological characters in Barneby and Grimes (1997, their Fig. 1) were not recovered by our analysis, whereas a node supported by seven morphological characters was resolved in our phylogeny (but with fewer sampled taxa) by the grouping of P. lanceolatum and P. winzerlingii. Even some clades supported by only one or two morphological characters in Barneby and Grimes (1996), such as a clade grouping Havardia sonorae and H. campylacantha and another clade grouping H. albicans, H. mexicana and H. pallens, were recovered in our analysis. However, one significant area of conflict involved the strongly supported sister relationship (PP = 1.0) between Pithecellobium and Sphinga in our study. In contrast, Painteria was resolved as the sister genus to Pithecellobium in Barneby and Grimes (1996), with four morphological characters supporting the result.

In cases of conflict between our molecular results and the morphological analyses of Barneby and Grimes (1996, 1997), we favour the former results for several reasons. First, although they mapped character state transitions on their phylogenies, Barneby and Grimes (1996) did not provide bootstrap values or other statistical measures of branch support in their phylogenies. Second, the morphological characters that were included in their phylogenetic studies included features, such as degree of pod compression, pod texture, pod curvature, valve reflection and coiling and degree development of an ovary stipe, which show continuous (or near continuous) variation when viewed across the entire Pithecellobium clade; thus, the division of the features into discrete states is subjective. Third, relative to molecular sequence data, suites of morphological characters may be more prone to homoplasy caused by shared selection pressures and/or developmental constraints (Scotland et al. 2003; Wiens 2004). Finally, relative to the phylogenies of Barneby and Grimes (1996), the well-supported phylogenetic results of our analyses show a greater degree of congruence with the results of previous molecular phylogenetic studies, including those based on analysis of other DNA regions (e.g. Brown et al. 2008; LPWG 2017; Koenen et al. 2020; Soares et al. 2021).

Other phylogenetic analysis in relation to the Pithecellobium clade sampled 15 species for matK (LPGW 2017) and 11 species for 997 nuclear genes (Ringelberg et al. 2022). They show different topologies for the Pithecellobium clade as well as the current topology (Fig. 3). The reasons for these differences could be due to the difference in the number of terminals, the number of nucleotides, alignment and analysis methods. Although the LPGW (2017) analysis has a few more terminals than Ringelberg et al. (2022), the latter has more key species (e.g. more taxa of Havardia and Painteria) and although this last study supports different generic relationships, it is congruent with the results presented here in demonstrating the non-monophyletic nature of Havardia and Painteria.

Our results further substantiate that three of the five genera of the Pithecellobium clade, Ebenopsis, Pithecellobium and Sphinga, are monophyletic, whereas Havardia and Painteria are not. While there exist multiple taxonomic solutions that would yield a classification consisting of only monophyletic genera (see Backlund and Bremer 1998; Humphreys and Linder 2009), we strongly favour an option that preserves the first three genera. Beyond minimising nomenclatural changes, such a solution is desirable since the first three genera are each morphologically distinctive and easily diagnosed.

There are two other nomenclatural options, which, in our opinion, are less appropriate. The first one would be to include all members of clade C (Fig. 3) - p.p. (excluding the type), Ebenopsis and Painteria - in a single genus named Ebenopsis, which has nomenclatural priority. However, this would result is a heterogeneous genus with respect to fruit morphology (Figs 2b, f and n), habit and ecology. The second option would be to recognise members of clade H (Fig. 3) as a new genus (as we do here) to transfer Painteria elachistophylla and P. revoluta to Ebenopsis. This new circumscription of Ebenopsis reduces morphological variation, but retains notable differences in habit and ecology. While Ebenopsis, although less expansive than in the first option, nevertheless encompasses trees growing in lowlands, Painteria comprises shrubs growing in highlands of up 2600 m elevation.

As for Havardia and Painteria, the preferred option retains these generic names for the clades that include the generic type species, H. pallens and P. revoluta, respectively. It thus requires the erection of two new genera: one for the clade comprised of H. sonorae and H. campylacantha, another for the species P. leptophylla. These four genera (the re-defined Havardia s.s. and Painteria s.s. and the proposed new genera) are individually diagnoseable by morphology.

Figure 2. 

Morphological diversity of the genera of the Pithecellobium clade as circumscribed here. Ebenopsis ebano (Berland.) Barneby & J.W. Grimes A inflorescence B pod. Gretheria campylacantha (L. Rico & M. Sousa) Duno & Torke C bark. Pithecellobium excelsum (Kunth) Mart. D bark. Havardia albicans (Kunth) Britton & Rose E inflorescence. Painteria elachistophylla (A. Gray ex S. Watson) Britton & Rose F pod G seed. Pithecellobium dulce (Roxb.) Benth. H leaves. Pithecellobium winzerlingii Britton & Rose I inflorescence. Pithecellobium keyense Britton J inflorescence. Pithecellobium lanceolatum (Humb. & Bonpl. ex Willd.) Benth. K inflorescence. Pithecellobium unguis-cati (L.) Benth. L pod and seed. Pithecellobium lanceolatum (Humb. & Bonpl. ex Willd.) Benth. M pod and seed. Sphinga acatlensis (Benth.) Barneby & J.W. Grimes N branch, leaves and inflorescences O pod. Sphinga platyloba (Bertero ex DC.) Barneby & J.W. Grimes P leaves and inflorescence. Photos: A, B, I–K, P German Carnevali C, D, N, O Colin E. Hughes E, L Gustavo A. Romero F, G Pedro Najéra Quezada, https://www.naturalista.mx) H Peter Pedersen M Rodrigo Duno.

Figure 3. 

Results of majority rule consensus tree of the Bayesian analysis of the nuclear ribosomal ETS and ITS regions the clade Pithecellobium.

Havardia, as here circumscribed, can be diagnosed within the complex for having both sylleptical and proleptical shoots, inflorescences arising from long shoots, leaves with one pair of pinnae and flowers with recurved corolla lobes. The new genus that we name Gretheria (containing H. sonorae and H. campylacantha) also has both sylleptical and proleptical shoots and inflorescences arising from long shoots, but the pinnae are distally accrescent and the corolla lobes are erect-ascending.

Painteria as redefined here, could be diagnosed by the combination of the valves of the pod elastically reflexed with age, the leaves with just one pair of pinnae and the corolla lobes erect-ascending, whereas the new monotypic genus Ricoa (comprising P. leptophylla), while also having the valves elastically reflexed with age, has leaves with the pinnae decrescent at each end and the corolla lobes recurved.

In the taxonomic treatment that follows, we provide descriptions (drawing heavily from data in Barneby and Grimes 1996) for the two new genera and make the necessary new combinations for the three species contained within them. Table 2 summarises the diagnostic characters of the seven recognized genera of the Pithecellobium clade.