1. Introduction

Dioscoreaceae comprises about 650 species and has a predominantly Pantropical distribution (Stevens 2001). Species of the family are distributed throughout Brazil, but especially in the Atlantic Forest and Cerrado biomes (BFG 2015). The family is currently accepted as monophyletic but with some uncertainties regarding its relationships with other families of the order Dioscoreales (Caddick et al. 2002), as well as with the position of the genus Tacca J.R. Forst. & G. Forst. (Merckx et al. 2006). The family comprises dioecious or monoecious volatile climbing or subshrub plants, most of which have an underground reserve organ (Couto et al. 2018). Their branches are aerial, unarmed, winged or aculeated, and the leaves possess petioles with a pulvinus at both ends. The staminate inflorescences are axillary or rarely terminal, isolated or clustered, pistillate or solitary. The fruits are generally of the capsule, berry or samara (Couto 2015). The family comprises the genera Stenomeris Planch (2 spp.), Trichopus Gaertn. (2 spp.), Tacca (17 spp.) and Dioscorea L. (633 spp. = 95% of the species of Dioscoreaceae), with Dioscorea representing most of the morphological, chemical and genetic diversity of the family (Govaerts et al. 2007).

Knuth (1924) treated Dioscorea as comprising four subgenera: Dioscorea subg. Helmia (Knuth) Uline (239 spp.); D. subg. Dioscorea Pax (337 spp.); D. subg. Stenophora (Uline) R. Knuth (15 spp.); and D. subg. Testudinaria (Salisb.) Uline (24 spp). The classification of Knuth (1924) included 38.8% of the species of the genus in D. subg. Helmia, which is characterized by the presence of an expanded wing toward the base of the seeds. The species of this subgenus are distributed throughout the tropical region, but most are in the Neotropics.

Wilkin et al. (2005), Viruel et al. (2016, 2018) and Couto et al. (2018) presented phylogenetic studies with species belonging to different subgenera and several of the sections of Knuth (1924). Few of the sections were found to be monophyletic, while other clades grouped species from several sections generally in concordance with geographic distribution, demonstrating a lack of phylogenetic support for the sections of Knuth (1924).

Couto et al. (2018) carried out the most comprehensive phylogenetic study to date with a greater representation of Neotropical taxa than previous studies. These authors resolved three New World clades – New World I (NW I), New World II (NW II) and New World III (NW III), with one isolated species in the clade New World IV (NW IV). Among these clades of Neotropical species, NW II groups the species of D. subg. Helmia, which occur in the Neotropics (the rest of the species of this subgenus are positioned so as to make it polyphyletic).

Couto et al. (2018) asserted that there is still a need for greater sampling of Neotropical clades to achieve a more resolved phylogeny. These authors also noted that the internal relationships of clades NW I and NW II have yet to be fully elucidated and still acknowledge the role of morphological characters, especially those of pollen, in understanding the relationships among the Neotropical species.

Palynological studies involving Dioscoreaceae in the works of Erdtman (1952), Su (1987), Caddick et al. (1998), Xifreda (2000), Schols et al. (2001), Schols, Wilkin et al. (2005), Wilkin et al. (2009) and Couto et al. (2014), verified the possible separation of species of Dioscorea, emphasizing the phylogenetic and taxonomic importance of the following characters: size, aperture number, exine ornamentation and the presence of orbicules.

There have been few palynological studies of this genus in the Neotropical Region. Barroso et al. (1974) studied both the taxonomy and the palynology of species of Dioscoreaceae occurring in the municipality of Rio de Janeiro and reported that the pollen grains are alike in form and aperture, yet may vary in the structure of the exine and the appearance of the surface. Recent studies have shown that the species Dioscorea campestris Griseb. and Dioscorea pedalis (Uline ex R. Knuth) R. Couto & J.M.A. Braga can be separated based on pollen grain size and sexine ornamentation (Couto et al. 2014).

In view of the above, new studies are needed on the pollen morphology of species of Dioscorea in the New Word clades NW I, NW III and, especially, NW II, to increase knowledge regarding taxa occurring in the Neotropics. Thus, the present palynological study involved 19 species selected to encompass a broad representation of the morphological diversity of species of these main Neotropical clades, with the aim of identifying taxonomic characters useful for distinguishing species and contributing to a better understanding of the phylogenetic relationships among species of the New World clades NW I, NW II and NW III.

2. Material and methods

Nineteen Neotropical taxa of Dioscorea were studied palynologically, with a focus on clade NW II. One specimen of each species was chosen for statistical treatment and illustration, and is indicated by an asterisk (^*) after the collector's name in the Supplemental material. Samples were obtained from fertile anthers of pre-anthesis flowers and/or buds from exsiccates deposited in the following herbaria: ESA, HB, NCU, R, RB and RFA (abbreviations in accordance with the Index Herbariorum; Thiers 2018, continuously updated).

The species were selected to account for a wide range of the morphological variability of the genus in the Neotropical Region, as well as its distribution in different habitats (Figure 1). The 19 studied taxa of Dioscorea cover all neotropical clades presented in the latest, and the majority of, phylogenetic studies of the genus (Viruel et al. 2016, 2018; Couto et al. 2018).

For analysis by light microscopy (LM), the material was processed according to the 60% lactic acetolysis method established by Raynal and Raynal (1971), because the pollen grains exhibited little resistance to traditional acetolysis. Pollen grains were photomicrographed under LM with a Canon PowerShot G6 digital camera coupled to a Zeiss Axiostar Plus binocular microscope with a 100× objective.

For analysis by scanning electron microscopy (SEM), anthers were separated from herbal material under a stereo-microscope, with the aid of sterilized forceps and stylets, and macerated. The non-acetolyzed pollen grains were then deposited on properly numbered aluminum supports covered with carbon tape (Melhem et al. 2003). The assembly was transferred to a vacuum pump, metalized with a layer of pure gold for about three minutes and examined under a JEOL JSM 6510 SEM located in the Laboratory of Images in Optical and Scanning Microscopy (LABIM) at the Federal University of Rio de Janeiro (UFRJ).

Randomly selected pollen grains in equatorial and lateral equatorial views were measured from at least three slides to homogenize the sample (Salgado-Labouriau 1973). The following measurements were made under a binocular microscope with an ocular micrometer: 25 measurements of maximum diameter (MD) and minimum diameter (mD) in polar view (PV); 10 measurements of polar diameter (PD) and equatorial diameter (ED), in equatorial view (EV); 10 measurements of MD and mD in lateral equatorial view (LEV); 10 measurements of the layers of the exine; and 10 measurements of the aperture (length and width). Classes were established for sulcus length as follows: short = 10.0–14.9 µm; long = 15.0–20.0 µm; and very long = 20.1–31.0 µm.

Ten pollen grains were measured in polar view for comparative purposes. A minimum of three permanent slides of acetolyzed pollen grains was assembled for each specimen to standardize the sample (Salgado-Labouriau et al. 1965). The pollen grains were measured within three days after preparation to avoid any changes in size (Salgado-Labouriau 1973; Wanderley and Melhem 1991). Types of ornamentation of the exine were analyzed and are described later. The slides used in the present study were deposited in the Palynotheca of the Laboratory of Palynology Álvaro Xavier Moreira, of the Department of Botany of the National Museum, Federal University of Rio de Janeiro.

Monocotyledonous pollen grains typically have a single aperture at the distal pole. The majority of species of the present study were found to have disulcate pollen grains. The pollen grains were preferentially positioned in polar view on the slides, making it difficult to obtain pollen grains in equatorial and lateral equatorial views. Pollen grains were recorded in four positions on the slides, as follows (Figure 2): (a) distal polar view: apertures located in the center of the pole; (b) proximal polar view: apertures located near the peripheries of the pollen grain; (c) equatorial view: apertures are partially seen in one of the peripheries of the pollen grain; and (d) lateral equatorial view: the two apertures are visible laterally in equatorial view.

A specific statistical treatment was performed on the results of the samples with 25 measurements, where the following were calculated: arithmetic mean (xā), standard deviation of the mean (s_x), 95% confidence interval (95% CI) and range of variation. The results of the statistical treatment and the morphopollinic analyses are presented in the tables.

Figure 1.

Maps of the distribution of Dioscorea species among different habitats. The distribution map was made using ArcGIS 10.5 (ESRI 2011), based on localities indicated in voucher specimens and further records collected and filtered from GBIF (2019), using the Biome2 vegetation layer from WWF (World Wide Fund for Nature, available in  http://services.arcgis.com/BG6nSlhZSAWtExvp/arcgis/rest/services/Biomes2/FeatureServer).

The terminology used in the pollen descriptions follows Erdtman (1952), Punt et al. (2007) and Hesse et al. (2009), who used the following size classes: small, 10–25 µm; and medium, 25–50 µm.

An exploratory analysis of the data was carried out with the aid of PC-ORD version 5.31 (McCune and Mefford 2011). Twenty pollen attributes (variables) of the analyzed species were organized in a matrix. Species names were abbreviated in graphs using the first four letters of the specific epithet while pollen attributes (variables) were expressed as letters.

The following variables (indicated by their respective abbreviations) were included in the multivariate analyses: small – SML; medium – MED; sulcus long – SL; sulcus very long – SVL; two apertures – AP2; one and two apertures – AP1-2; margin – M; operculum – OPER; rugulate-microreticulate – RUG-MICR; reticulate-cristate – RET-CRIS; perforate – PERF; rugulate – RUG; rugulate-reticulate – RUG-RET; rugulate-perforate – RUG-PERF; spiculate – ESP; striate – ESTR; vermiculate-perforate – VER0-PERF; vermiculate – VER; and reticulate – RET; microreticulate – MICRORRET.

Qualitative data were expressed in the matrix as ‘1' for presence and ‘0’ for absence.

Principal component analysis (PCA) was used to order the variables and determine whether species were grouped by pollen attributes. The variance-covariance (var-cov) was obtained from the means of the morphometric data in the palynological analysis and ordered in a biplot chart based on Euclidean distance. The results are shown in a two-dimensional graph with the first and second principal components. The values of the vectors for each axis and the total cumulative variance are represented in the tables, as is the character matrix for cluster analysis.

Hierarchical cluster analysis (cluster-HCA) was performed to classify the analyzed species into groups that share (similarity) pollen variables. Two aspects were considered in relation to the groups formed and the set of analyzed variables: the percentage of information (variables) needed to form the groups, and the final number of groups formed. A dendrogram was constructed from the cluster analysis using Euclidean distance (Caccavari et al. 2008) and Ward's linkage.

Figure 2.

Position of the pollen grains.

Figure 3.

Cluster analysis performed with the measured variables of pollen from species of Dioscorea L. acant - D. acanthogene; anoma - D. anomala; campe - D. campestris; campo - D. campos-portoi; chond - D. chondrocarpa; convo - D. convolvulacea; coron - D. coronata; demou - D. demourae; gland - D. glandulosa; grand - D. grandiflora; grise - D. grisebachii; molli - D. mollis; monad - D. monadelpha; multi - D. multiflora; ovata - D. ovata; plani - D. planistipulosa; sinco - D. sincorensis; stege - D. stegelmanniana; subha - D. subhastata

3. Results

The following characteristics were used in the morphological analysis of pollen grains of Dioscorea (Plates 1–4; Figures 2–4; Tables 1–4): dispersion unit, polarity, contour, size, number and type of aperture(s), exine ornamentation and operculum presence/absence.

3.2. Size

Pollen grains of the analyzed species of Dioscorea varied from small (nine species) to medium (10 species). In polar view (Table 2) the smallest maximum diameter was for D. campestris (20.5 µm), while the greatest maximum diameter was for D. anomala (35.5 µm). Pollen fallen in equatorial view was found for only four species: D. campestris, D. glandulosa, D. monadelpha and D. ovata; D. campestris had the smallest while D. campestris had the largest (Table 3). In lateral equatorial view, D. campestris had the smallest maximum diameter while D. multiflora had the greatest (Table 4). Measurement of the maximum diameter in polar view revealed differences in the ends of the pollen grains, which were classified as the following contours (shapes): truncate for 16 taxa (D. acanthogene, D. chondrocarpa, D. anomala, D. campos-portoi, D. coronata, D. demourae, D. glandulosa, D. grandiflora, D. grisebachii, D. mollis, D. multiflora, D. ovata, D. planistipulosa, D. sincorensis, D. stegelmanniana and D. subhastata); acute-truncate for two taxa (D. convolvulacea and D. monadelpha – Plate 1, figure 10; Plate 3, figure 4); and acute for one taxon (D. campestris –Plate 1, figure 5).

Plate 1.

Photomicrographs and electromicrographs of Dioscorea species: D. acanthogene – 1. polar view: general aspect (SEM), 2. detail of the surface (SEM). D. anomala – 3. polar view: optical section (LM), 4. general aspect (SEM). D. campestris – 5. polar view: general aspect and aperture (SEM), 6. detail of the surface (SEM). D. campos-portoi – 7. polar view: general aspect (LM). D. chondrocarpa – 8. polar view: general aspect (SEM), 9. detail of the surface (SEM). D. convolvulacea – 10. polar view: general aspect (SEM), 11. detail of the ends of the aperture (SEM), 12. detail of the surface (MEV). Scale bars: 1, 3–5, 7, 8, 10 = 5 µm; 2, 6, 9, 11, 12 = 1 µm.

Plate 2.

Photomicrographs and electromicrographs of Dioscorea species: D. coronata – 1. polar view: general aspect (SEM), 2. detail of the surface (MEV). D. demourae – 3. polar view: outline (LM), 4. equatorial lateral view: optical section (LM), 5. polar view: general aspect (SEM), 6. detail of the surface (SEM). D. glandulosa – 7. polar view: general aspect and one aperture (SEM), 8. general aspect and two aperture (SEM), 9. detail of the surface (MEV). D. grandiflora – 10. polar view: general aspect and aperture (SEM). D. grisebachii – 11. polar view: aperture (LM), 12. general aspect and aperture (SEM), 13. surface (LM). Scale bars: 1, 3–5, 7, 8, 10–13 = 5 µm; 2, 6, 9 = 1 µm.

Plate 3.

Photomicrographs and electromicrographs of Dioscorea species: D. mollis – 1. polar view: outline (LM), 2. detail of the surface (SEM). D. monadelpha – 3. polar view: aperture (SEM), 4. polar view: general aspect (SEM), 5. equatorial lateral view: general aspect and surface (SEM). D. multiflora – 6. polar view: aperture (LM), 7. equatorial lateral view: optical section (LM), 8. polar view: general aspect and surface (SEM). D. olfersiana – 9. equatorial lateral view: general aspect and ends of the aperture (SEM). 10. detail of the surface and ends of the aperture (SEM). D. ovata – 11. polar view: outline (LM), 12. aperture (LM), 13. general aspect (SEM), 14. detail of the surface (SEM). Scale bars: 1, 4, 6–13 = 5 µm; 2, 3, 14 = 1 µm; 5 = 2 µm.

Plate 4.

Photomicrographs and electromicrographs of Dioscorea species: D. perdicum – 1. polar view: general aspect and aperture (SEM), 2. general aspect: surface (SEM). D. planistipulosa – 3. polar view: aperture (LM). D. sincorensis – 4. polar view: general aspect and aperture (SEM), 5. detail of the surface (SEM). D. sinuata – 6. lateral view: outline (LM), 7. polar view: general aspect and surface (SEM). D. stegelmanniana – 8. polar view: general aspect (SEM), 9. detail of the surface (SEM). D. subhastata – 10. polar view: general aspect and surface (SEM). D. therezopolensis – 11. polar view: general aspect and aperture (SEM), 12. detail of the surface (SEM). Scale bars: 1, 2, 4, 6, 8, 10, 11 = 5 µm; 3 = 10 µm; 5, 9, 12 = 1 µm; 7 = 2 µm.

Figure 4.

Principal component analysis (PCA) performed with the measured variables of pollen from species of Dioscorea L.

3.5. Multivariate analysis

3.5.1. Hierarchical cluster analysis (HCA)

Relationships among the analyzed species were depicted with a dendrogram that explained about 42.32% of the data. Four clusters were formed when 50% of the remaining information (variables) was analyzed (Figure 3):

• - Cluster I: D. acanthogene, D. campos-portoi, D. sincorensis and D. campestris.

• - Cluster II: D. convolvulacea, D. ovata, D. subhastata, D. monadelpha, D. stegelmanniana and D. demourae.

• - Cluster III: D. chondrocarpa, D. coronata, D. grandiflora, D. grisebachii, D. multiflora, D. anomala, D. mollis and D. planistipulosa.

• - Cluster IV: D. glandulosa.

3.5.2. Principal component analysis (PCA)

The PCA was significant and explained 59.35% of the total variance. The first two axes (components) were significant and explained 44.87% and 14.48% of the variation, respectively. The significant variables for axis 1 were disulcate aperture, long sulci, small size and reticulate ornamentation of the sexine. The significant variables for axis 2 were very long sulcus, monosulcate and disulcate apertures, small size and reticulate ornamentation of the sexine (Figure 4).

The species are fairly well dispersed between the negative and positive poles of axes 1 and 2, with few apparent groupings because there are few characteristics that differentiate these species.

The species D. planistipulosa, D. mollis, D. grandiflora, D. chondrocarpa, D. grisebachii and D. coronata are located toward the positive pole of axis 1 because they possess medium-sized pollen grains with a very long sulcus. The species D. glandulosa and D. anomala are positioned near each other and distanced from the others because they possess pollen grains with both monosulcate and disulcate apertures, and D. glandulosa has an operculum. The species D. campestris, D. sincorensis, D. campos-portoi and D. acanthogene are close to each other at the negative pole of axis 1 because they have small pollen grains with a long sulcus. The species D. campestris is distanced from the others by possessing one and two apertures.

The species D. glandulosa, D. anomala and D. campestris were differentiated toward the positive pole of axis 2 based on having monosulcate and disulcate pollen. Among these, D. campestris differs in having a small size, revealing that this character is very significant in the analysis of similarity of the pollen grains. The species D. convolvulacea and D. ovata are similar; D. subhastata and D. monadelpha are close to each other, differing only in the ornamentation of the exine; and the species D. stegelmanniana and D. demourae share only a disulcate aperture. These six species are located on the same pole in both axes. At the negative pole of axis 2 and the positive pole of axis 1, the species D. chondrocarpa, D. coronata, D. grandiflora and D. grisebachii are differentiated only by the ornamentation of the sexine; D. planistipulosa and D. mollis are similar to each other, and D. multiflora is isolated from the others because of the medium size of its pollen and its very long sulcus and reticulate sexine.

Table 1.

Morphologic characters of Dioscorea pollen grains.

Table 2.

Measures (in µm) of pollen grains in polar view of Dioscorea species (n = 25): larger diameter (LD) and smaller diameter (SD). x – arithmetic mean; sx – standard deviation of the mean; C.I. – confidence interval.

4. Discussion

Palynological analysis of 19 taxa of Dioscorea revealed important pollinic attributes that contribute to its delimitation, such as size, contour (shape), number of apertures and sexine ornamentation.

Pollen grains with a fine exine, as found for the species of Disocorea analyzed here, do not allow the traditional acetolytic treatment of Erdtman (1952), and only the methodology for fragile or very fragile pollen grains of Raynal and Raynal (1971) could be used. This confirmation of exine fragility allows the inference to be made that it would be difficult to preserve these pollen grains in their natural habitat.

Table 3.

Measures (in µm) of pollen grains in equatorial view of Dioscorea species (n = 10): polar diameter (PD) and equatorial diameter (ED). x – arithmetic mean.

The studied species of Dioscorea exhibited morphometric variation with both small and medium-sized pollen grains. This variation was also observed by Erdtman (1952), Su (1987), Schols et al. (2001, 2003), Schols, Wilkin et al. (2005), Wilkin et al. (2009) and Couto et al. (2014). Large pollen grains were recorded for species from Africa (D. buchananni Benth, D. nipponica Makino) and the species D. pyrenaica Bubani & Bordère ex Gren of Europe (Su 1987; Schols et al. 2001; Schols, Wilkin et al. 2005).

Barroso et al. (1974) studied both the palynology and the taxonomy of 24 species of Dioscorea and affirmed that the pollen grains are equal in shape and aperture. The results obtained here revealed a difference in the number of apertures because D. glandulosa exhibited a pattern of 1(-2) sulcate. The above authors did not record monosulcate pollen grains, which may have been due to their low percentage in the slides analyzed.

Table 4.

Measures (in µm) of pollen grains in equatorial lateral view of Dioscorea species (n = 10): larger diameter (LD) and smaller diameter (SD). x – arithmetic mean.

Table 5.

Measures (in µm) of the apertures and layers of the exine of pollen grains of the Dioscorea species (n = 10).

Su (1987) described the pollen morphology of 33 species of Dioscorea distributed among five sections and emphasized variation in the number of apertures, with the majority of the pollen grains being disulcate, with the exception of Stenophora, which had one aperture, and Opsophyton, which had two or three. Other authors reported the occurrence of monosulcate and disulcate pollen grains, including Caddick et al. (1998), Schols et al. (2001), Schols et al. (2003), Schols, Furness et al. (2005), Schols, Wilkin et al. (2005), Wilkin et al. (2009) and Viruel et al. (2010). The results obtained here corroborate those of the aforementioned authors but disagree with Su (1987) who recorded trisulcate pollen grains.

Couto et al. (2014) analyzed the pollen morphology of D. campestris and D. pedalis (Uline ex R. Knuth) R. Couto & J.M.A. Braga, but did not observe disulcate pollen grains. The palynological analysis of the present study found that D. campestris varied in the number of openings, with both monosulcate (7%) and disulcate (93%) pollen grains.

Barroso et al. (1974) divided species of Dioscorea into groups based on the ornamentation of the sexine: (A) reticulate-ornamented, (B) reticulate, (C) ornamented and (D) striated. Of the 26 species analyzed by these authors, six are included in the present study, none of which had pollen grains with ornamentation that can be classified into the groups of Barroso et al. (1974). This divergence in results is possibly due to the previous authors' lack of resources related to SEM, which brings clarity regarding the type of sexine ornamentation.

Zavada (1983), Caddick et al. (1998) and Schols et al. (2001) described different patterns of sexine ornamentation: finely reticulate, reticulate, striate, gemmate and rugulate-reticulate. In addition to the ornamentation patterns defined by theses authors, nine more were recorded in the present study: perforate, microreticulate, reticulate-cristate, spiculate, rugulate, rugulate-perforate, rugulate-microreticulate, vermiculate and vermiculate-perforate. However, the present study did not find gemmate or finely reticulate patterns. These differences, according to Couto et al. (2018), may be because the species they studied were not Neotropical and, therefore, are among different clades than are the species studied here.

Schols et al. (2003) and Schols, Wilkin et al. (2005) analyzed 61 species of Dioscorea, of which 54% were perforate, 28% were striate, 6% varied from perforate to microreticulate (characteristics of both ornamentations are observed in the same pollen grain), 6% were cerebroid-perforate, 3% were rugulate, 2% were gemmate and 1% were perforate to rugulate. The taxa analyzed here exhibited similar characteristics of pollen grain ornamentation.

Couto et al. (2014) palynologically analyzed D. campestris and D. pedalis and concluded that there are differences in details of the sexine (lumina with or without ornamentation). The present study corroborated these findings of Couto et al. (2014).

According to Fraga (2016), the species D. piperifolia Humb & Bompl. ex Will., D. glandulosa and D. grandiflora are part of a taxonomic complex of critical species, differentiated only by a few macromorphological characters and exhibiting little variation. Couto et al. (2018) considered D. piperifolia and D. grandiflora to form a clade with strong support in all their analyses. The pollinic attributes recorded here revealed a distinction between these two species, with D. grandiflora having only disulcate pollen grains with spiculate ornamentation while D. glandulosa has monosulcate and disulcate pollen grains in the same specimens, and is the only species with an operculum and rugulate ornamentation, microreticulate with the presence of granules, showing that palynology did not corroborate the suggested phylogeny.

According to Tenorio et al. (2017), the species D. mollis, D. chondrocarpa and D. planistipulosa are anatomically similar to species of the Old World with two phloem units between great vessels of meta-xylem, while taxa of the clades NW I and NW II possess only one unit of phloem. Palynologically, D. mollis and D. planistipulosa grouped very close in the hierarchical cluster analysis due to possessing a microreticulate sexine, which differs from the perforate sexine of D. chondrocarpa.

The hierarchical cluster analysis formed four large groups. Cluster I contained D. acanthogene, D. campos-portoi and D. sincorensis, which were all very similar, and D. campestris, which was separated from the others, which were only disulcate, by being 1–2-sulcate.

Cluster II contained: D. convolvulacea and D. ovata, which were very similar; D. subhastata and D. monadelpha, which differed only in ornamentation; and D. stegelmanniana and D. demourae, which were linked by a single character (disulcate aperture). Cluster III was subdivided into two groups, with D. mollis and D. planistipulosa being very similar.

Cluster IV contained only D. glandulosa, which was differentiated by having 1–2 apertures and the presence of an operculum, the latter standing out in comparison to the other species.

The PCA revealed that the species located at the negative pole of axis 1 were those of clusters I and II, while those at the positive pole belonged to clusters III and IV. The species D. convolvulacea, D. ovata, D. monadelpha and D. subhastata, belonging to Cluster I (at negative poles of axes 1 and 2), were similar due to the strong characteristic of small size, along with being disulcate and having long sulci. The three species D. campestris, D. anomala and D. glandulosa were distinguished because they had pollen grains with both aperture patterns (monosulcate and disulcate), which was supported by their location at the positive pole of axis 2.

The pollinic attributes analyzed here partially corroborate the phylogenetic analysis of Couto et al. (2018) since the pollen data in the multivariate analysis separated D. subhastata from D. monadelpha, while in the phylogeny they were in the same clade. This divergence may be due to the fact that microcharacters, including those of pollen, were not used in the phylogenetic analysis.

5. Conclusion

The results obtained by the present work allowed the analyzed species to be organized into six groups based on sexine ornamentation: reticulate, microreticulate, perforate, vermiculate, rugulate and spiculate.

Fifty-three percent of the studied species had their pollen grains characterized for the first time: D. acanthogene, D. campos-portoi, D. convolvulacea, D. demourae, D. grandiflora, D. grisebachii, D. monadelpha, D. planistipulosa, D. sincorensis and D. stegelmanniana.

Some pollen grains of species of Dioscorea (D. anomala, D. campestris and D. glandulosa) were characterized as having both monosulcate and disulcate aperture patterns in the same specimen.

The present study did not corroborate the sections of the taxonomic classification of Knuth (1924), indicating that a reassessment of established subgenera is needed, as well as a better understanding of phylogenetic relationships. Both the principal component analysis and the cluster analysis confirmed this result, since species of different sections were grouped in the same clusters and dispersed on axes 1 and 2.