Monopteryx is a florally divergent genus of Dipterygeae, an early-branching papilionoid legume clade largely marked by winged papilionate floral architecture, expanded upper calyx lobes often assuming a wing-shaped orientation and petals differentiated into standard, wings, and a keel enclosing the basally connate stamens. In contrast to the remaining Dipterygeae genera, Monopteryx has differentiated petals but the marginally coherent keel with interlaced trichomes exposes the free stamens and the expanded calyx upper lobes are nearly entirely fused with a standard-like dorsal orientation. Monopteryx species are restricted to the Amazonian rainforests, where they have diversified since the last ~15 Ma, but the divergence of the genus is estimated to be as old as ~39 Ma. They grow as large buttressed trees usually with a uniquely “flying” architecture, which are arched from the trunk to the ground and separated from one another, unlike that found in any other species of leguminaceous trees. Its fruits are elastically dehiscent pods, and, in some species, they bear marginally crimped wings along the sutures. Our taxonomic revision of this ecologically and evolutionarily important, ancient genus includes an analysis of about 135 specimens from across 14 herbaria, including both type and historical collections, as well as recently collected samples from our extensive fieldwork across remote areas of the Amazon. Grounded on a densely-sampled dated molecular phylogeny of nuclear and plastid data, here we recognize three phylogenetically and morphologically distinct taxa: M. angustifolia, M. inpae, and M. uaucu. After a careful revision of their nomenclatural history, we also found that M. inpae was not validly published. We subsequently have provided typification of all names associated with species of the genus. This revision also includes morphological descriptions, illustrations, and distribution maps for all species. We also discuss the phylogenetic relationships between the species and the evolution of selected taxonomically key morphological characters in the context of the entire Diptergyeae clade.

In order to assess the phylogenetic distinctiveness and relationships of all Monopteryx species, a new phylogenetic tree of the Dipterygeae was generated based on nuclear ribosomal ITS/5.8S and the plastid trnL intron and protein-coding matK DNA sequences. The new Dipterygeae phylogeny has greatly augmented the sampling of multiple accessions per species in Monopteryx after sequencing herbarium specimens and our newly collected samples preserved in silica gel. Comprehensive sampling across Monopteryx and related Dipterygeae genera was possible due to previously published taxonomically curated, vouchered sequences (Cardoso et al. 2012b, 2015b; Carvalho et al. 2023a) available in GenBank (http://www.ncbi.nlm.nih.gov/genbank/).

The DNA isolation, polymerase chain reaction (PCR), and amplifications follow the same protocols as described in our previous molecular phylogenetic studies (Cardoso et al. 2012b; Queiroz et al. 2015; Carvalho et al. 2023a). The products of the forward and reverse sequencing reactions were analyzed on ABI3730XL sequencers (Applied Biosystems) following the manufacturer’s protocol at the sequencing facility of the Rede de Plataformas Tecnológicas in FIOCRUZ-Bahia. The sequenced reads were assembled into contigs with Geneious (Drummond et al. 2009) and saved as a fasta file. All newly generated sequences and those retrieved from GenBank were aligned manually in SeaView v. 4 (Gouy et al. 2010). Gaps in sequences of the protein-coding matK gene were inserted by comparing the amino-acid-translated sequences (Wojciechowski et al. 2004).

We ran a combined Bayesian analysis of phylogenetic reconstruction in MrBayes v. 3.2.1 (Ronquist et al. 2012) via the CIPRES Science Gateway v.3.3 on-line portal (www.phylo.org) (Miller et al. 2010). A best-fit nucleotide substitution model was first selected via the Akaike information criterion (AIC) in jmodeltest2 (Darriba et al. 2012). The best fit-models were GTR+I+G for the ITS/5.8S dataset, GTR +G for the matK, and GTR +G for the trnL intron. Eight simultaneous chains were initiated with randomly drawn trees and run 10 million generations in two separate runs of a Metropolis-coupled Markov Chain Monte Carlo (MCMC) permutation of parameters, sampling one tree every 10,000th generation. Non-correlated samples at the stationary phase were summarized in a 50% Bayesian majority-rule consensus tree after a burn-in of 25%. Posterior probabilities (PP) represent support measures (Huelsenbeck et al. 2002), where we considered the following thresholds for node support: weak between 0.5–0.94 and strong above 0.95. Visualization and partial editing of the Bayesian tree for graphical presentation were done in FigTree v.1.4.4 (Rambaut 2012).

The combined dataset (ITS/5.8S, matK, and trnL intron) was used to estimate the molecular divergence times through the Bayesian uncorrelated lognormal relaxed-clock model (Drummond et al. 2006) in order to discuss the coalescence of the Monopteryx species in the context of a rainforest-associated evolutionary history (Pennington and Lavin 2016). The analysis was implemented in BEAST package v.1.8.2 (Drummond et al. 2012), via the CIPRES Science Gateway and incorporated the same substitution models used in the phylogenetic reconstruction, a random starting tree, and a Yule speciation process. Two fossil-calibrated nodes with lognormal prior age distributions (see Ho 2007) and one secondary calibration with normal prior distribution (Lavin et al. 2005) were chosen to obtain absolute ages. The fossil flowers of Barnebyanthus that are found in the USA (Crepet and Herendeen 1992; Herendeen and Wing 2001) were used to calibrate the root at 55 Mya (offset = 55.0 mean = 0.0 and stdev = 1.0). The crown node of the Swartzieae (sensu Cardoso et al. 2013b) was calibrated at 45 Mya (offset = 45.0, mean = 0.0 and stdev = 1.0) based on fossil fruits and leaves of the southeastern USA suggesting an affinity with Swartzia (Herendeen 1992). The estimated ages of Lavin et al. (2005) were used to calibrate the ADA clade. The priors for the parameter ucld mean gamma were shape = 0.001 and scale = 1000. The BEAST running file was generated in BEAUti v.1.8.2 (Drummond et al. 2012), by enforcing Dipterygeae and each of its constituent genera to be monophyletic, as strongly supported by the Bayesian combined analysis. Two independent MCMC of 100 million generations were run, sampling parameters every 10000 generations after a 10% burn-in period. Tracer v.1.6 (Rambaut and Drummond 2013) was used to check the convergence and stationarity, and all parameter estimates had ESS (effective sample size) values >200. LogCombiner was used to combine independent runs, and a maximum clade credibility (MCC) tree was generated using the TreeAnnotator. The MCC tree was annotated as a chronogram with median ages and 95% highest posterior density (HPD) intervals of node ages, and visualized in FigTree v.1.4.4.