Introduction

Lysiloma Benth. (^{Bentham, 1844}) is one of 36 genera within the tribe Ingeae (^{Brown, 2008}) in the extended subfamily Caesalpinioideae (^{LPWG, 2017}). Two recent phylogenetic analyses (^{Brown et al., 2008}; ^{Iganci et al., 2015}) suggest that Lysiloma and the monotypic Hesperalbizia Barneby & J.W. Grimes are sister taxa. As currently understood, it is composed of eight tree species with a primarily Neotropical distribution, ranging from Arizona and Florida through Mexico, Central America, and the Antilles (^{Thompson, 1980}; ^{Gale and Pennington, 2004}; ^{Andrade and Sousa, 2012}) (Table 1).

Lysiloma acapulcense (Kunth) Benth. and L. divaricatum (Jacq.) J.F. Macbr. are the species with the widest distribution, ranging from northern Mexico to Central America (Nicaragua and Costa Rica). Both taxa include a long list of synonyms and broad morphological variation. These taxa urgently require a deeper future reassessment. The rest of the species have more restricted distributions (Table 1). The general distribution of the genus coincides in general terms with Megamexico 3, a biogeographic area that includes the entire Mexican territory, but also the southern part of the United States of America (the northern portions of the Chihuahuan and Sonoran deserts in California, Arizona, New Mexico, and Texas) in the north, and Central American to Nicaragua north of the lakes, in the south (^{Rzedowski, 1991}). Two exceptions are L. latisiliquum (L.) Benth., and L. sabicu Benth., which are found in Florida, the Bahamas, and the West Indies (^{Thompson, 1980}). In general, Lysiloma species occupy a wide variety of dry forest types, being dominant or codominant elements in many of them (^{Rascón-Ayala et al. 2018}; ^{Ancona et al., 2019}; ^{Ortiz-Ávila et al., 2020}).

The generic name of this group of plants derives from the Greek Lysis (to lose) and loma (edge), referring to the shedding of the legume edge at fruit maturity that allows for the dispersion of the seeds. This type of dehiscence has been referred to as craspedial (^{Thompson, 1980}; ^{Gale and Pennington, 2004}). Two distinctive species of the genus lack this feature, Lysiloma sabicu and L. latisiliquum (Fig. 1E, F), where the valves remain together at maturity. An additional, distinctive feature of the genus is the membranous, foliaceous stipule (Fig. 1C) (^{Bentham, 1875}), a character state that is shared by Hesperalbizia. ^{Thompson (1980)} proposed an informal classification system with two subgenera based upon fruit dehiscence characters: Lysiloma subg. Lysiloma Thompson (nom. nud.), with indehiscent or late dehiscent legumes and Lysiloma subg. Lysivalva Thompson (nom. nud.), where the legume is dehiscent, and the valves break apart at the margins. Subgenus Lysiloma is composed of two sections defined by inflorescence types (capitulum vs. raceme) and flower pedicels (sessile vs. pedicellate): Lysiloma subg. Lysivalva sect. Capitata Thompson (nom. nud.), and sect. Racemosa Thompson (nom. nud.) (Fig. 1A, B, D).

Figure 1: General morphology of Lysiloma Benth. A. leaves, and inflorescences of Lysiloma acapulcense (Kunth) Benth.; B. inflorescence of L. acapulcense; C. leave and stipule of Lysiloma latisiliquum (L.) Benth.; D. inflorescence of Lysiloma latisiliquum; E. young and old fruits of L. latisiliquum; F. fruit of Lysiloma sabicu Benth. Pictures A, B: Claudia Ramírez; C, D, E: Rodrigo Duno de Stefano; F: Susan Ford Collins. 

As is the case of most members of tribe Ingeae, Lysiloma pollen grains are arranged in polyads of 16 (28-32 in L. divaricatum (Jacq.) J.F. Macbr.) (^{Guinet and Grimes, 1997}). In general, polyads are 46-79 µm diameter, isodiametric where the eight radial pollen grains are longer than wide, whereas the internal tetrad is composed of square pollen grains. The external pollen grains are dissymmetric in thickness; whereas the tectum is perforated with 20-100/µm² ornamentation is uniform throughout, fossulate, or polygonal with more or less rounded areoles (^{Sorsa, 1969}; Guinet and Grimes, 1997). Hesperalbizia occidentalis (Brandegee) Barneby & J.W. Grimes is characterized by similar yet larger, symmetrical polyads. Both Lysiloma and Hesperalbizia feature 2n=26 chromosomes (^{Thompson, 1980}; Rico ^{Arce, 1992}).

Lysiloma is known from a Tertiary (Olygocene) fossil record, L. mixtecana^{Magallón-Puebla & Cevallos-Ferriz (Magallón-Puebla and Cevallos-Ferriz, 1993}). This fossil was recovered in a sedimentary layer in Puebla, along with records of other members of tribe Ingeae: Pithecellobium grimesii Calvillo-Canadell & Cevallos-Ferriz and Pithecellobium barneby Calvillo-Canadell & Cevallos-Ferriz (^{Calvillo-Canadell and Cevallos-Ferriz, 2005}).

Assessing the monophyly of Lysiloma is particularly relevant since two related genera studied from a molecular perspective were non-monophyletic: Abarema Pittier (^{Iganci et al., 2015}), and Zygia P. Brown (^{Ferm et al., 2019}). Aiming at testing the monophyly and sister-group relationships of Lysiloma, we performed phylogenetic analyses combining morphological and molecular datasets. We employed morphological characters, plastid (trnK and matK), and nuclear (ETS) DNA sequence data.

These analyses were carried out with parsimony for the morphological characters and parsimony and Bayesian inference for the molecular and total evidence approach. This study will allow us to test three hypotheses: 1) Lysiloma is monophyletic as currently circumscribed; 2) Lysiloma and Hesperalbizia are closely related (^{Brown et al., 2008}; ^{Iganci et al., 2015}), and 3) the infrageneric relationships between the species as proposed by ^{Thompson (1980)}. Regarding infrageneric relationships, the informal classification proposal of Thompson (1980) was used as a null hypothesis. Moreover, we were interested in estimating the chronology of the diversification under the uncorrelated lognormal relaxed molecular clock approach. To do so, we adopt the same assumption of ^{Becerra (2005)} regarding Bursera Jacq. ex L. (Burseraceae) and the expansion of the dry forest. In other words, the history and evolution of Bursera in Mexico mirrors the history of the dry forest in Mexico. Thus, the very forces that drove the range expansions and contractions of the dry forests also shaped the diversification of Bursera. Because Lysiloma is highly adapted to the ecological conditions of the dry forest, we test whether the diversification of this genus is related to the expansion of the dry forest in the Miocene (20-5 Mya (million years ago)), when arid environments expanded across the world.

Material and Methods

Results

Morphology

The parsimony analysis based on morphological characters yielded eight most parsimonious trees (L=63, CI=69, RI=80). Figure 2 shows the strict consensus tree, where three nodes have no support and collapse. Vachellia farnesiana, upon which the cladograms are rooted, is followed by a poorly supported grade that includes Senegalia parviceps, Mariosousa dolichostachya, Acaciella spp. plus a clade containing Lysiloma and Hesperalbizia. The clade including Acaciella Britton & Rose as a sister group of Lysiloma and Hesperalbizia has moderate bootstrap support (77%) and is held together by four synapomorphies. Acaciella has a strong bootstrap support (97%) and is supported by two synapomorphies. The clade Hesperalbizia and Lysiloma (clade A) also features a high bootstrap support (91%) and is supported by four synapomorphies. Hesperalbizia is supported by one synapomorphy. On the other hand, Lysiloma (clade B) is monophyletic and has a low bootstrap support (68%); this clade features five morphological synapomorphies. Internally, Lysiloma is poorly resolved; there are three terminal taxa; L. auritum (Schltdl.) Benth., L. acapulcense and L. watsonii Rose and clade C. The last clade lacks bootstrap support, but it is held together by four reversions. It is composed of L. latisiliquum and two clades. The first, including L. divaricatum + L. microphyllum Benth. with a low bootstrap support (53%), and the second composed of three species with moderate bootstrap supported (77%): L. candidum Brandegee as a sister group of L. tergeminum Benth., and L. sabicu.

Clade C, which is composed by L. watsonii and the clade L. acapulcense and L. auritum is strongly supported by the bootstrap (94%). Clades D (L. candidum plus the clade L. sabicu + L. tergeminum), E (L. sabicu + L. tergeminum), and F (L. divaricatum + L. microphyllum) lacked significant bootstrap support (<50%).

Figure 2: Results of parsimony analysis of the genus Lysiloma Benth. based on morphology (value below the node, bootstrap support, simbols: A, clade Hesperalbizia and Lysiloma; B, clade Lysiloma; C, species with inflorescence racemose; black triangle (u), synapomorphy: stipule developed, black circle (˜) and black square (¢): synapomorphies for inflorescence racemose and pod craspedial; white circle (™): inflorescence capitate (reversion), and white square (£): pod indehiscent (reversion). 

Molecular data

Figure 3 shows the result of the Bayesian analysis (rDNA-ETS). In this analysis, some of the genera segregated from Acacia Mill. such as Acaciella, Mariosousa Seigler & Ebinger, Senegalia Raf., and Vachellia Wight & Arn. are poorly represented. However, this large analysis evaluates the relationship of the genus Lysiloma in a more general context (see ^{Brown et al., 2008}, for a complete list of the genera of the Acacieae+Ingeae tribes). Acaciella, Senegalia, Mariosousa comprise a basal grade. The clade A (PP=1.0) is formed for two clade: Calliandra is the sister group Zapoteca and both are the sister group of Ingeae, including Acacia (B Clade, PP=1.0). This clade has two subclades: the clade Viguieranthus sensu ^{Rodriguez de Souza et al. (2016)} including Faidherbia A. Chev., Sanjappa É.R. Souza & M.V. Krishnaraj, Thailentadopsis Kosterm., and Viguieranthus Villiers (PP=0.95) and clade C. This clade included Cojoba Britton & Rose as sister group of Lysiloma and Hesperalbizia, and clade D (PP=0.99). This last clade includes four subclades whose relationships are unresolved, namely the New World Ingeae p.p. clade, which is strongly supported (PP=1.0), the poorly supported (PP=0.76) Old World Ingeae clade, including Acacia, and lastly, the Pithecellobium clade (part of the New World Ingeae), which is strongly supported (PP=1.0).

Figure 3: Results of Bayesian analysis of the genus Lysiloma Benth. based on a large molecular matrix of DNA-ETS (Clade A: core Ingeae, including Acacia Mill. clade B: clade A- clade Calliandra, and Zapoteca; clade C: Cojoba, sister group of Lysiloma and Hesperalbizia, and Ingeae p.p. including Acacia, Clade D: New World Ingeae p.p., Old World Ingeae and Acacia, and Pithecellobium alliance. Important synapomorphies: stipule developed for Lysiloma + Hesperalbizia, and pod craspedial for Lysiloma. 

Combined molecular and total evidence analyses

From all regions analyzed, we found ETS to be the most informative: 26 taxa, 482 characters, from which 44.73% were informative, followed by trnK, 21 taxa, 999 characters, 5.4% informative, and by matK, 18 taxa, 767 characters, 3.65% informative. The results of the combined molecular analysis and those of the total evidence analysis are very similar; thus, we only show the second in Figure 4. At the base of the tree is a polytomy conformed by Vachellia farnesiana (root), Acaciella, and then it follows a clade with all the other taxa (Clade A). This polytomy is the sister group of a grade with Senegalia parviceps, Mariosousa dolischotachya, and clade B. This last clade included four subclades without resolution among them, namely Calliandra spp., Faidherbia, Zapoteca spp., and finally a clade with Lysiloma and Hesperalbizia (Clade C). In the combined molecular analysis as well as each individual analysis (ETS, matK, trnK), there is no resolution for Hesperalbizia occidentalis, Lysiloma sabicu and the core of Lysiloma. However, in the total evidence analysis, Hesperalbizia was recovered as sister group of Lysiloma with low support.

Figure 4: Results of total Bayesian analysis the genus Lysiloma Benth. (morphology, ETS, matK and trnK). * PP> 95 %. 

Diversification of the genus Lysiloma

The exclusion of the Lysiloma fossil from the molecular clock analyses does not result in relevant differences in chronology. Hence, the results are not contingent upon the restriction imposed by the inclusion of this fossil and associated date.

Figure 5 shows the result molecular clock analyses. The divergence of Lysiloma and Hesperalbizia (L1) was estimated at about 28.22 myr (95% HPD: 19.5-36.25) in the beginning of the Oligocene. The second node (L2) corresponds to the divergence of Lysiloma and was estimated at about 29.86 myr (95% HPD: 28.64-39.13). Node L3, the divergence of the Lysiloma core, was estimated at about 22.65 myr (95% HPD: 15.78-28.33) in the beginning of the Miocene.

Figure 5: Chronogram of Lysiloma, and Hesperalbizia and other related taxa from the tribe Acacieae and Ingeae (Caesalpinioideae) based on ETS data. Divergence times are shown using the computer program BEAST v. 1.10. The calibration nodes 1 (28.4 myr), and 2 (16 myr), are marked by dots based on the fossil records. The root of the tree was set to no more than 45 myr. 

Discussion

Phylogeny reconstruction is a crucial aim of contemporary systematics. The success of phylogenetic inference can be measured in terms of resolution, support, and accuracy (^{Wortley et al., 2005}). Many studies suggest that increasing sequence data is a better way to improve support, resolution, and accuracy of the phylogenetic trees (^{Rokas and Carroll, 2005}) at any level; the ordinal level (^{Li et al., 2019}), family level (^{Stull et al., 2015}), generic level (^{Cardoso et al., 2015}; ^{Stull et al., 2015}), and at the species level (^{Nicholls et al., 2015}). We consider that additional DNA regions must be explored to test and improve the resolution of topology, particularly regarding Lysiloma and Hesperalbizia. The morphological analysis (maximum parsimony) confirms the monophyly of Lysiloma and its close relationship with Hesperalbizia, in concordance with two previous molecular analyses, which include at least one accession of each genus (^{Brown et al., 2008}; ^{Iganci et al., 2015}). However, the current morphological analysis does not allow to recover a complete picture of the relationships of Lysiloma and Hesperalbizia with other members of tribes Acacieae and Ingeae, because only a few taxa were included.

Molecular and total evidence analyses

In a more general context, three previous phylogenetic analyses have been published with abundant accessions of ITS, and/or ETS (^{Brown et al., 2008}; ^{Rodriguez de Souza et al., 2013}; ^{Iganci et al., 2015}). In concordance with those, our large (rDNA-ETS, Fig. 2) analysis supports a basal position for the segregates of Acacia, namely Acaciella, Mariosousa, Senegalia, and Vachellia as well as for the Lysiloma+Hesperalbizia clade. It also supports the monophyly of the Pithecellobium alliance and the clade Viguieranthus as a sister group of Zapoteca (^{Rodriguez de Souza et al., 2016}). None of these earlier phylogenetic analyses support the relationship of Lysiloma and Hesperalbizia found by ^{Grimes (1995)}, nor support the relationship of Hesperalbizia with Albizia (^{Rico Arce et al., 2008}). In the current analysis some clades are strongly supported (e.g. Zapoteca, Cojoba, and the Pithecellobium alliance (PP >98%)), but not the Lysiloma+Hesperalbizia clade. The basal clade of the current tree (Fig. 2) is a mixture of elements restricted to the New World (Acaciella, Mariosousa, and Zapoteca), others from the Old Word (Faidherbia), as well as elements from both regions (Calliandra+Afrocalliandra É.R. Souza & L.P. Queiroz, and Senegalia). There are three clades with no resolution which reveal geographic coherence: The New World Ingeae (e.g. Abarema, Inga Mill., Samanea, Enterolobium Mart., etc.), the Old World Ingeae (e.g. Archidendron F. Muell., Paraserianthes I.C. Nielsen, Wallaceodendron Koord., etc.), plus Acacia s.s., and the Pithecellobium alliance from the New World.

Regarding the internal relationship of the genus, Lysiloma latisiliquum and L. sabicu, the two taxa bearing fruits with persistent lateral sutures are not retrieved as a clade; instead, the two species are nested in widely diverging clades, pointing to an independent evolution of this fruit type. Our results suggest that the distribution of L. sabicu in the Antilles can be explained by an old vicariance process but in the case of L. latisiliquum, it requires long-distance dispersal (from mainland Mexico to south Florida, Bahamas, Turks and Caicos Islands, Cuba and Haiti). A fact that still has no explanation is why L. sabicu has been less successful in its expansion in the Antilles compared to L. latisiliquum, considering the long occupation time of the lineage in the area.

Lysiloma taxa with capitate inflorescences and those with racemose inflorescences constitute polyphyletic assemblages. In summary, we found no support for ^{Thompson’s infrageneric informal classification (1980)}. This was already proposed by ^{Barneby and Grimes (1996}), who did not find any reason to maintain such a classification and proposed that character states observed in L. latisiliquum and L. sabicu probably evolved independently. The evolution of fruit dehiscence in Lysiloma is of ecological interest, because indehiscent fruits, which are buoyant and probably an apomorphic condition, have most likely allowed the genus to invade the West Indies.

In our analysis, we did not find a clear biogeographical pattern with the three species distributed in the Nearctic region (the southwestern coast of USA, and Mexico): L. candidum Brandegee, L. microphyllum (partially) and L. watsonii, because they are not closely related, suggesting the genus invaded the Nearctic region thrice. However, a structurally and geographically coherent clade recovered in both the morphological and total evidence analyses was found to occur along the Pacific coast of Mexico: L. candidum, L. tergeminum, and L. divaricatum. One of the morphological features of this clade are the leaves with few pinnae. Therefore, diversification in Lysiloma seems to have followed an isolation by allopatry pattern, whereby sister taxa are currently allopatric and sympatric taxa are not closely related. Thus, Lysiloma latisiliquum and L. sabicu, which are sympatric in the West Indies, belong to different clades in the genus, whereas the sister pair of L. watsonii and L. auritum occur one each on one of the opposing drainages of Mexico, the former on the Pacific slope, the latter on the Gulf slope.

Lysiloma sabicu is the most distinct taxon within the genus. This taxon shares with Hesperalbizia leaves with few pinnae with a few large leaflets, which can imply a morphological transition. It is possible that additional evidence will eventually increase the support for a sister group relationship of Hesperalbizia and Lysiloma. In the total evidence analysis (Bayesian), Hesperalbizia is the sister group of Lysiloma, but the relationship is poorly supported (PP <0.95%).

The advantages of including Hesperalbizia in the genus Lysiloma are similar to those of creating a monotypic genus for Lysiloma sabicu. Any decision here, regarding the generic boundaries of Lysiloma and Hesperalbizia, should be well thought-out and meet the mandatory rule of monophyly, as well as the secondary criteria proposed by ^{Backlund and Bremer (1998)}: support, diagnosability, maximum informativity, and stability. There are morphological differences in the most distinctive character of Lysiloma as compared to Hesperalbizia; for example, the fruit in Lysiloma exfoliates, eventually revealing the pale brown endocarp (not in Hesperalbizia). Another difference is related to flower number: fewer with longer petals in Hesperalbizia. Seeds are also different (lens-like, pale brown, and areolate in Hesperalbizia), as opposed to ovate, oblong to elliptic and dark brown to black in Lysiloma. Thus, here we argue that until more data become available, it is best to retain Lysiloma in its present circumscription.

Despite Lysiloma being a small genus, it shows a complex evolutionary structure. The phylogenetic molecular analysis points at the homoplasic evolution of many morphological characters, which strongly suggest that different lineages were modeled by similar ecological pressures. Other examples of a similar pattern are the lizards of the genus Anolis Daudin, where in the case of Caribbean species, the evidence strongly supports the hypothesis of repeated and independent development of similarly shaped body on each island (^{Losos, 2001}). In the case of flowering plants, the genus Manihot Mill. is another good example of convergent evolution (in this case, growth forms) (^{Cervantes-Alcayde et al., 2015}). Both examples are of speciose lineages with 400 and 100 species, respectively. Because of this phenomenon (repeated evolutionary convergence), the classifications based only on morphology may be misleading, being not natural (monophyletic), but polyphyletic or paraphyletic. Moreover, although these two examples may suggest that repeated evolutionary convergence is characteristic of diverse lineages, our data of Lysiloma show that even small genera with more restricted distributions can also show this pattern. A similar pattern of evolution could be found in other members of the tribe Ingeae, involving some Old World Albizia species and New World Albizia species associated with flooded forests, all the species of the genus Hydrochorea Barneby & J.W. Grimes, and the two species of Balizia Barneby & J.W. Grimes section Balizia. All these lineages are subject to environmental pressures (seasonally flooded environments) and present indehiscent, lomentiform fruit.

In addition, there is a long-standing argument on the position of Lysiloma microphyllum relative to L. divaricatum, to whose synonymy it has been relegated. ^{Thompson (1980)} recognized both as distinct, albeit morphologically very similar. However, ^{Barneby and Grimes (1996)} and ^{Gale and Pennington (2004)} were unable to identify consistent characters to separate them. Upon evaluating morphological features in herbarium specimens, we did not find differences in character states between both taxonomic concepts. Furthermore, molecular data do not support both taxa as separate entities, as their ETS, matK, and trnK sequences are almost identical. Neither is there a geographical discontinuity in their distributional ranges (Thompson, 1980). Thus, here we conclude that L. divaricatum and L. microphyllum should be treated as the same species (see ^{Gale and Pennington, 2004}; ^{Andrade and Sousa, 2012}).

Lysiloma acapulcense requires special consideration. As presently circumscribed, this taxon includes a long list of synonyms, reflecting the fact that it is an extremely polymorphic and widely distributed species. A review of the fruit of the types allows us to recognize at least two additional morphotypes: one characterized by very long and thin fruits (Lysiloma jorullense Britton & Rose), and a second morph bearing fruits with a broad base (Lysiloma platycarpa Britton & Rose). Thus, this species requires a taxonomic re-evaluation.

Diversification of the genus Lysiloma

The divergence between Lysiloma and Hesperalbizia (Fig. 5, L1) was estimated at 28.2 myr at the beginning of the Oligocene (Fig. 5, Table 4). This date coincides with a drastic and abrupt decrease in temperature and humidity at the beginning of the Oligocene (^{Galeotti et al., 2016}). The second node (L2) corresponds to the divergence of the Lysiloma estimated at 26 myr, coinciding with two important facts, a low global temperature and humidity as well as a proximity between the Caribbean Arc and the American continent through the Yucatan peninsula during the Oligocene (^{Pindel and Kennan, 2009}). An important lineage divergence occurred in the upper Miocene, whereas additional minor ones happened below this node (N3) all along the Miocene, which agree with ^{Becerra’s (2005)} findings regarding the evolution and divergence of Bursera (Burseraceae) in the dry forests.

We can assume that the history of Lysiloma seems to be tightly related to the last 30 to 5 myr corresponding with the formation of the main orographic systems in Mexico, such as the Sierra Madre Occidental and later the Neovolcanic belt and the dry forest (^{Gómez-Tuena et al., 2007}). The last uplift of the Sierra Madre Occidental occurred between 34 and 15 myr, whereas the Neovolcanic axis was formed in several stages along a west-east progression that started in the west (^{Sosa et al., 2018}). The uplift of these two orographic systems was presumably responsible for the climatic conditions necessary (high temperature and low humidity) for the development and maintenance of the dry forests (^{Sosa et al., 2018}).

Conclusion

Lysiloma and Hesperalbizia are sister genera, albeit with low clade support, with Lysiloma often resolved as paraphyletic with respect to Hesperalbizia. More data are necessary to confirm the generic status of Lysiloma. This sister clade is early branching within the clade comprising the rest of the tribe Ingeae. Indeed, we estimate the stem age of this sister clade to average about 32 Ma. In addition, our analysis shows that, although Lysiloma is a small genus, it shows a complex evolutionary structure that may be modeled in different lineages of the genus by the same ecological pressures, suggesting that the phenomena of convergence and/or parallel evolution occur regardless of species richness. One consequence of this phenomenon (repeated evolutionary convergence) is that classifications based on morphology are not necessarily natural (monophyletic), but instead, a conglomerate of polyphyletic or paraphyletic taxa. This applies not only to morphology, but also to patterns of geographic distribution: unrelated species can occupy the same distribution.

Regarding the divergence time, the history of the genus Lysiloma and Hesperalbizia begins in the Oligocene and diversifies into the beginning of the Miocene. This time frame coincides with low global temperatures, a proximity between the Caribbean Arc and the American continent through the Yucatan peninsula, and the uplift of the Sierra Madre Occidental and later of the Neovolcanic axis. Presumably, these conditions were responsible for the climatic conditions associated with the development and maintenance of the dry forest in Mexico (^{Sosa et al., 2018}), the habitat in which Lysiloma diversified.