A total of 13 species of Amorimia were sampled (out of 15), comprising both subgenera currently recognised by Almeida et al. (2017a) (Supplementary material 1). Only two species of Amorimia could not be sampled (A. andersonii R.F.Almeida and A. tumida R.F.Almeida & A.C.Marques) due to the lack of flower buds or flowering specimens. Ectopopterys soejartoi W.R.Anderson and Mascagnia cordifolia (A.Juss.) Griseb. were sampled as outgroup (Supplementary material 1).

For light microscopy (LM), the pollen grains were treated with the acetolysis method (Erdtman 1960), with modifications cited by Melhem et al. (2003). The obtained microscope slides were incorporated into the pollen slide collection from the Plant Morphology and Palynology Lab, São Paulo State University, Campus Jaboticabal. Pollen micrographs were taken using an optic microscope Leica IM50 coupled with a video camera and computer, and digitally treated and edited using Photoshop. For scanning electron microscopy (SEM), the material was prepared following Melhem et al. (2003) for non-acetolised pollen grains. The pollen electron micrographs were generated using a JEOL JSM5410 scanning electron microscope.

Pollen descriptions follow the terminology by Barth and Melhem (1988), Punt et al. (2007), and Bellonzi et al. (2020) for LM. The pollen grains of the analysed Amorimia species using SEM are described following Halbritter et al. (2018). Shape and size terminology follows Erdtman (1952), and exine thickness follows the classification (i.e. very thin, thin, or thick) proposed by Faegri and Iversen (1950). Diameter measurements (DI and DII) were taken within seven days of acetolysis to avoid pollen grain alterations related to the method (Melhem and Matos 1972). All measurements were randomly taken for 25 pollen grains. In contrast, other pollen characters (i.e. aperture, total exine, sexine, nexine, and tectum) were randomly measured for only ten pollen grains (Melhem and Matos 1972; Salgado-Laboriau 1973). The measurements referring to the tectum are included in the value of the sexine, as this structure is part of the sexine.

We calculated the arithmetic mean (x), average standard deviation (sx), sample standard deviation (s), coefficient of variability (CV), and 95% confidence interval following Zar (1996) and Vieira (2011). Only the arithmetic average (x) was calculated for measurements of n = 10. Diameter I and II measurements were used to compare diameter values of the analysed pollen grains for Amorimia and outgroups using MINITAB v.14 (Zar 1996; Vieira 2011). One principal component analysis (PCA) was performed using FITOPAC 1 (Shepherd 1996) and PC-ORD 5 (McCune and Mefford 2011) to verify the influence of pollen grain quantitative data on species ordination and grouping. The analysis used nine metric variables: diameter I (DIAI), diameter II (DIAII), ectoaperture length (ECLEN), ectoaperture width (ECWID), exine (EXIN), sexine (SEXI), nexine (NEXI), tectum (TECT), and shape (SHAP) (Supplementary material 2).

The consensus phylogenetic tree by Almeida (2018) was pruned with Mesquite v.2.73 (Maddison and Maddison 2006) to show only the outgroup sampled in our study and used for the character mapping of the pollen morphology. Character coding followed the recommendations of Sereno (2007) for morphological analyses. Primary homology hypotheses (De Pinna 1991) were proposed for pollen shape, size, ornamentation, exine structure, and apertures. A total of 12 pollen characters were scored for Amorimia and outgroups (Supplementary material 3). In addition to Mascagnia cordifolia, we sampled M. sepium from the palynological literature (Makino 1986) for the character mapping analyses. All characters were optimised on the concatenated tree using the Maximum Likelihood function (mk1 model) using Mesquite v.2.73 (Maddison and Maddison 2006) and visualised with Winclada (beta) v.0.9 (Nixon 1999).

Pollen grains of all studied species of Amorimia, Ectopopterys, and Mascagnia are monads (Figs 1–4), polar in Ectopopterys or apolar, medium (Figs 1–2, 4–5; Supplementary material 4) to large (Figs 3–5), with circular amb, oblate-spheroidal to prolate-spheroidal (Figs 1–4). Apertures zonocolporate, 3-colporate, with long and narrow ectoaperture and circular endoaperture in Ectopopterys (Fig. 3F–G) or pantoporate 6-porate in Amorimia, and 8-porate with colpoids (Figs 1–4) in Mascagnia (Fig. 3H–I). Exine tectate, sexine rugulate (Fig. 2C–D), with or without areolae (Fig. 1L) or psilate (Fig. 1B) regions near the pores or distributed over the pollen grain surface (Fig. 1G–H). The exine thickness varies from very thin (M. cordifolia) to thin or thick (Supplementary material 5) according to the averages of the diameters 1 and 2, with sexine thicker than the nexine (Fig. 3F; Supplementary material 5).

Pollen grains are monads, apolar, medium or large (Amorimia velutina and A. kariniana), oblate-spheroidal or prolate-spheroidal, 6-porate, pantoporate, with colpoids (sometimes not evident in A. amazonica) (Supplementary material 4). Exine tectate, sexine rugulate, with psilate regions (only in A. candidae, A. velutina, A. amazonica, and A. septentrionalis; Figs 1A–C, 2C–F, 3C–E, 4A–D, N–O, Q–R) or with areolate regions. The exine thickness varies from thin to thick (A. maritima, A. rigida, A. velutina, A. camporum, and A. concinna) and sexine is thicker than nexine (Supplementary material 5). In SEM, the pollen grains are clypeate (except for A. amazonica; Fig. 4A–B) and the exine fossulate (A. amazonica, A. pubiflora, A. velutina; Fig. 4A–B, N–P) or psilate-perforate (other species; Fig. 4C–G, I–M, Q–T).

Pollen grains are monads, polar, medium, circular amb, prolate-spheroidal, 3-colporate, zonocolporate, with long and narrow ectoaperture, without margo, and circular endoaperture. Exine tectate, sexine rugulate with areolate regions. The exine thickness is thin, sexine thicker than nexine (Supplementary material 5).

Pollen grains are monads, apolar, medium, prolate-spheroidal (Supplementary material 4), 8-porate, pantoporate, with not evident colpoids. Exine tectate, sexine rugulate with psilate regions. The exine thickness is very thin, sexine thicker than nexine (Supplementary material 5).

The quantitative data analyses used pollen grain diameters and their respective averages and confidence intervals (Supplementary materials 4–5). We observed three distinct groups when analysing the metric values of the diameters (Fig. 5): 1. Smallest diameter species (A. amazonica, A. camporum, and A. concinna), 2. Intermediate diameter species (A. coriacea, A. exotropica, A. maritima, A. rigida, A. pubiflora, A. septentrionalis, and M. cordifolia), and 3. Largest diameter species (A. candidae, A. kariniana, A. pellegrinii, A. velutina, and E. soejartoi).

The PCA summarised 78.82% of the total variability of the data, in which axis 1 was more informative to the PCA since it summarised 43.47% of the variability (Fig. 6). The analysed Mascagnia and Ectopopterys were recovered far from those of Amorimia (Fig. 6). The species of A. subg. Uncinae showed lower values for all metric variables analysed (negative side of axis 1), except for A. camporum and A. kariniana, which were positioned with the species of A. subg. Amorimia. For axis 1, the most significant variable was ECLEN (Supplementary material 2), which distinguished Ectopopterys from other species by the colporate ectoaperture. Axis 2 summarised 31.34 % of the variability in our data; the most significant variables for this axis were SEXI and ECWID (Supplementary material 2). Note that the species of Mascagnia and Ectopopterys had lower values for these variables and were close to A. amazonica. In general, the species of A. subg. Uncinae are positioned on the positive side of axis 2, except for A. camporum, which appears alongside the species of A. subg. Amorimia with higher values for the variables that stand out in the ordination of axis 2.

All lineages from the molecular phylogeny were recovered with at least one or more homoplasies/apomorphies, except for both Amorimia subgenera (A. subg. Amorimia and A. subg. Uncinae). Ectopopterys soejartoi was recovered supported by three homoplasies regarding exine thickness (3.00–3.99 µm), tectum thickness (1.00–1.50 µm), and aperture width (4.00–4.99 µm), and a single synapomorphy regarding the exine ornamentation (rugulate with areolate regions). The Amorimia + Mascagnia clade was recovered supported by a single homoplasy regarding the aperture length (6.00–6.99 µm) and two synapomorphies regarding the apertures type (porate) and aperture number (= 6). Mascagnia was recovered supported by one homoplasy regarding sexine thickness (1.00–1.99 µm) and two synapomorphies regarding the number of apertures (= 8) and exine thickness being very thin. Mascagnia cordifolia was recovered supported by a single synapomorphy regarding the aperture length (3.00–3.99 µm), while M. sepium was recovered supported by a single homoplasy regarding aperture width (5.00–5.99 µm) and two synapomorphies regarding exine thickness (5.00–5.99 µm) and nexine thickness (3.00–3.99 µm) (Fig. 7, Table 1, Supplementary material 3).

Amorimia was recovered supported by a single homoplasy regarding nexine thickness (0.01–0.99 µm) and three synapomorphies regarding exine thickness (4.00–4.99 µm), aperture width (6.00–6.99 µm), and exine ornamentation (psilate-perforate). Both subgenera of Amorimia were not recovered, supported by any homoplasy or synapomorphy. Within A. subg. Amorimia, A. exotropica was recovered supported by two homoplasies regarding sexine thickness (3.00–3.99 µm) and aperture width (7.00–7.99 µm). The A. velutina + A. coriacea + A. maritima + A. candidae + A. pellegrinii + A. andersonii + A. rigida clade was recovered supported by three homoplasies regarding pollen grains shape (oblate-spheroidal), nexine thickness (1.00–1.99 µm), and aperture length (5.00–5.99 µm). Amorimia velutina was recovered supported by four homoplasies regarding pollen grain size (large), tectum thickness (1.51–1.99 µm), exine thickness (thick) and ornamentation type (rugulate with psilate regions), and two autapomorphies regarding exine thickness (6.00–6.99 µm) and sexine thickness (4.00–4.99 µm). The A. coriacea + A. maritima + A. candidae + A. pellegrinii + A. andersonii + A. rigida clade was supported by a single synapomorphy regarding aperture length (7.00–7.99 µm). Amorimia coriacea was recovered supported by three homoplasies regarding pollen grain shape (prolate-spheroidal), exine thickness (3.00–3.99 µm), and nexine thickness (0.01–0.99 µm). The A. maritima + A. candidae + A. pellegrinii + A. andersonii + A. rigida clade was supported by two homoplasies regarding tectum thickness (0.51–0.99 µm) and aperture width (7.00–7.99 µm). Amorimia maritima was recovered supported by two homoplasies regarding sexine thickness (3.00–3.99 µm) and exine thickness (thick). The A. candidae + A. pellegrinii + A. andersonii + A. rigida clade was supported by a single homoplasy regarding aperture length (6.00–6.99 µm). Amorimia candidae was recovered supported by four homoplasies regarding exine thickness (3.00–3.99 µm), nexine thickness (0.01–0.99 µm), tectum thickness (1.51–1.99 µm), and ornamentation type (rugulate with psilate regions). The A. pellegrinii + A. andersonii + A. rigida clade was recovered supported by a single homoplasy regarding aperture length (6.00–6.99 µm). Amorimia pellegrinii was recovered supported by a single homoplasy regarding pollen grain shape (prolate-spheroidal). The A. andersonii + A. rigida clade was recovered supported by three homoplasies regarding sexine thickness (3.00–3.99 µm), tectum thickness (0.51–0.99 µm), and exine thickness (thick) (Fig. 7, Table 1, Supplementary material 3).

Finally, within the Amorimia subg. Uncinae, A. tumida was not recovered as supported by any homoplasy or autapomorphy. The A. pubiflora + A. septentrionalis clade was recovered supported by two homoplasies regarding exine thickness (3.00–3.99 µm) and aperture length (5.00–5.99 µm). Amorimia pubiflora was recovered supported by a single homoplasy regarding exine ornamentation type (fossulate). Amorimia septentrionalis was recovered supported by three homoplasies regarding nexine thickness (1.00–1.99 µm), ornamentation type (rugulate with psilate regions), and aperture width (5.00–5.99 µm). The A. camporum + A. kariniana + A. amazonica + A. concinna clade was recovered supported by two homoplasies regarding sexine thickness (3.00–3.99 µm) and tectum thickness (1.00–1.50 µm). Amorimia camporum was recovered supported by three homoplasies regarding pollen grain shape (oblate-spheroidal), tectum thickness (0.51–0.99 µm), and exine thickness (thick), and two autapomorphies regarding aperture length (8.00–8.99 µm) and aperture width (9.00–9.99 µm). Amorimia kariniana was recovered supported by three homoplasies regarding pollen grain size (large), exine thickness (3.00–3.99 µm), and sexine thickness (2.00–2.99 µm). The A. amazonica + A. concinna clade was recovered as supported by a single synapomorphy regarding exine ornamentation (rugulate). Amorimia amazonica was recovered supported by three homoplasies regarding exine thickness (2.00–2.99 µm), sexine thickness (1.00–1.99 µm), nexine thickness (1.00–1.99 µm), and exine ornamentation type (fossulate), besides a single autapomorphy regarding tectum thickness (0.01–0.50 µm). Finally, Amorimia concinna was recovered supported by four homoplasies regarding pollen grain shape (oblate-spheroidal), aperture length (4.00–4.99 µm), aperture width (4.00–4.99 µm), and exine thickness (thick) (Fig. 7, Table 1, Supplementary material 3).

The genus Amorimia was recently segregated from Mascagnia, and, unfortunately, no palynological evidence was included in its original description (Anderson 2006). Lowrie (1982) performed a comprehensive study of the pollen morphology of 60 out of the 75 currently accepted genera of Malpighiaceae, describing three main morphological types and a few subtypes for this family. Radially symmetric pollen grains divided into 1. 3-colporate, 2. parasyntricolporate, 3. syntricolporate, 4. 4-colporate, or 5. polycolporate genera. Globally symmetric pollen grains lacking ectoapertures divided into 1. Aspidopteroid, 2. Bunchosioid, and 3. Ryssopteroid types, and globally symmetric pollen grains with ectoapertures divided into 1. Banisterioid, 2. Clonodioid, 3. Mascagnioid, and 4. Tetrapteroid types (Lowrie 1982). This author described the pollen grains of some species of Mascagnia, now treated in Amorimia, as of the Mascagnioid subtype, characterised as pollen grains with branched exine ornamentation with fused rugae and more than six pores randomly dispersed by the intersection of two rugae. Our results demonstrated clypeate pollen grains with fossulate or psilate-perforate exine for all species of Amorimia, different from those presented by Lowrie (1982). Amorimia pollen grains are similar to the pollen grains of the genus Mascagnia but differ concerning the number of apertures, corroborating the taxonomic changes proposed by Anderson (2006). In general, our data agree with the denomination of Mascagnioid pollen grains, as previously done by Lowrie (1982), for Amorimia species, since the pollen grains present patterns of aperture and ornamentation somewhat similar to those observed by Lowrie (1982), varying however in the number of apertures and ornamentation type, which allows to use them as diagnostic traits to easily distinguish Amorimia from Mascagnia.

Makino (1986) and Belonsi and Gasparino (2015) also found 8-porate pollen grains in Mascagnia cordifolia and M. sepium. Still, they observed nexine thicker than sexine, different from the data found in this study (i.e. sexine thicker than nexine). Therefore, the number and type of apertures corroborate the pollen pattern found for the species of Mascagnia, even with slight differences observed in the present study. The pollen grains of Ectopopterys soejartoi were briefly described by Anderson (1980) as 3-colporate, with the present study also corroborating the description presented by this author. Pollen grain type and number of apertures are also useful to differentiate Amorimia species from their allied genera (Ectopopterys and Mascagnia).

Another factor to be highlighted is the corroboration obtained from the ancestral pollen character reconstructions that placed colporate pollen grains (as in Ectopopterys soejartoi) as the probable ancestral state and porate pollen grains (Amorimia and Mascagnia) as a derived character in the Malpighioid clade. We corroborated that the pollen morphology of Amorimia regarding qualitative characters is constant for the species analysed, but showed that quantitative characters are very informative for their taxonomy. According to traditional palynological classification, Amorimia can be considered stenopalynous (i.e. with minor discrete morphological variations). As abovementioned, the type and number of apertures allow the distinction of Amorimia from the closely related Mascagnia and Ectopopterys. In the PCA analysis, the metric variables of the pollen grains confirm the qualitative data and help to distinguish the analysed genera since the differences in the measurements of the ectoapertures and the exine layers allowed Mascagnia and Ectopopterys to be separated from Amorimia.

In LM, the pollen grains of the species analysed here present sexine rugulate with areolate or psilate areas, which was also verified in previous studies for Malpighiaceae (Makino 1986; Makino-Watanabe et al. 1993a, 1993b, 1998; Belonsi and Gasparino 2015). For the description of pollen ornamentation in SEM, clypeate pollen grains were observed (as described by Halbritter et al. 2018), and the sexine is fossulate or psilate-perforate, details not previously described for species of Amorimia.

Cameron et al. (2001) tested the phylogenetic relevance of Lowrie’s (1982) three main micromorphological pollen groups in Malpighiaceae. These authors found that radially symmetrical pollen grains are probably plesiomorphic in Malpighiaceae, occurring in the Byrsonimoid, Acridocarpoid, Mcvaughioid, and Ptilochaetoid clades. Since this author did not sample any outgroups outside Malpighiaceae in his study, it is impossible to confidently state the plesiomorphic nature of this pollen grain morphology in Malpighiaceae. Nonetheless, Perveen and Quaiser (1995) show that the pollen grains of two species of Bergia L. (Elatinaceae) are indeed radially symmetrical, corroborating this character state as probably plesiomorphic for the Elatinaceae + Malpighiaceae clade. The same pattern of pollen morphology was also reported for two species of Elatine L. (Ramayya and Rajagopal 1971), increasing the chances of radially symmetrical pollen grains being, indeed, a synapomorphy for the clade formed by both families. Additional studies for the remaining 54 accepted species in Elatinaceae are needed for an in-depth assessment of this hypothesis.

In contrast, globally symmetrical pollen grains with ectoapertures were recovered as a synapomorphy for the Bunchosioid + Hiraeoid + Tetrapteroid + Malpighioid + Stigmaphylloid clade by Cameron et al. (2001). Globally symmetrical pollen grains without ectoapertures were recovered as homoplastic, having independently arisen at least seven times mostly in Old World genera of Malpighiaceae (except for the New World Barnebya W.R.Anderson & B.Gates, Bunchosia Rich. ex Kunth, and Heladena A.Juss.; Cameron et al. 2001). Amorimia and allies have constantly been placed by several molecular phylogenetic studies (Cameron et al. 2001; Davis et al. 2001, 2014; Davis and Anderson 2010; Almeida et al. 2017a; Almeida 2018) as early diverging lineages in the Malpighioid clade, showing globally symmetrical pollen grains with ectoapertures, as described by Lowrie (1982). The remaining lineages of the Malpighioid clade are endemic to the Old World (i.e. Africa + Asia) and show globally symmetrical pollen grains without ectoapertures as an adaptation to the generalist pollination syndrome arisen in Old World lineages due to the lack of oil collecting bees in this region of the planet (Lobreau 1968; Lowrie 1982; Cameron et al. 2001; Davis et al. 2014).

Our results recovered two evolutionary patterns regarding homoplastic and apomorphic pollen micromorphological characters. Regarding synapomorphies or autapomorphies (i.e. apomorphies), few qualitative characters, such as the type of pollen aperture (colporate or porate) and the number of apertures (3, 6, or 8), were very informative in distinguishing lineages at the generic rank. A few quantitative characters, such as exine thickness and aperture width, were informative at the generic level for the species sampled in this study. On the other hand, a few quantitative characters such as exine, nexine, sexine, and tectum thickness, and aperture length and width were informative to distinguish lineages at an infrageneric rank. Only the ornamentation type was informative to distinguish species at an infrageneric level. Regarding homoplasies, all quantitative and almost all qualitative micromorphological pollen characters analysed were informative both at the generic and infrageneric levels in Amorimia and allies. It is also worth mentioning that most of the homoplastic characters recovered as informative in our results were related to quantitative pollen characters, which are frequently underexplored in evolutionary studies of pollen grains.

Finally, Almeida et al. (2017a) proposed two subgenera within Amorimia: A. subg. Amorimia, or Atlantic clade, and A. subg. Uncinae, or Amazonian clade. As the clade names suggest, these species are already geographically separated. Molecular parsimony and bootstrap analyses showed that this separation is supported, and some morphological characters also support both subgenera. The authors suggest a differentiation of both subgenera based on pollen characters such as pollen amb and size (Almeida et al. 2017b). Nonetheless, the data from the present study do not corroborate Almeida et al. (2017b), since their species sampling for Amorimia was incomplete. In fact, when this classification was proposed, Amorimia tumida was still unknown to science and was not included in the molecular phylogeny published by these authors (Almeida et al. 2017b). This species is sister to the remaining species of the A. subg. Uncinae and was described only based on molecular data, and vegetative and fruit morphology, with its flowers still being unknown to science (Almeida et al. 2017b). Consequently, it was impossible to analyse this species’ pollen morphology in the present study. Since A. tumida is a crucial species placed at the base of the A. subg. Uncinae clade, this taxon’s missing data directly impacts the phylogenetic reconstruction of the analysed pollen micromorphological characters. Thus, only future studies focusing on the pollen morphology of A. tumida will be able to shed light on the relevance of pollen morphology to corroborate the classification system of Amorimia.