Comparative Pollen Morphology of the Genus Chaenomeles Lindl. (Rosaceae): Diagnostic Features and Implications for Taxonomy

: The taxonomic placements of Chaenomeles


Introduction
The genus Chaenomeles Lindley belongs to the tribe Maleae of the family Rosaceae and is a small group of trees and shrubs distributed in temperate areas of East Asia [1][2][3]. Chaenomeles species have long been widely appreciated as ornamentals in Asia and Europe, and their fruits, commonly known as Mugua in China, have demonstrated considerable potential in the food and pharmaceutical industries [4][5][6]. According to the classifications of Flora of China [2], it comprises five wild diploid (2n = 34) species, which are C. sinensis (Dum.Cours.) Koehne, C. cathayensis (Hemsl.) Schneider, C. japonica (Thunb.) Lindl., C. speciosa (Sweet) Nakai and C. thibetica Yü [2,3,7]. Interspecific hybrids are of common occurrence in all possible combinations between three of the species (C. cathayensis, C. japonica and C. speciosa), and hybrid taxa produced from artificial breeding have consequently been described [1,8]. Molecular and morphological evidence of spontaneous interspecific Therefore, the present study was designed with the aims (1) to document and illustrate the morphology of optimally preserved pollen materials of the taxonomically problematic genus Chaenomeles using both light and scanning electron microscopy; (2) to identify informative (qualitative and quantitative) pollen features of taxonomical significance; (3) to categorize Chaenomeles taxa based on statistical analysis (PCA and cluster analysis), and evaluate whether the results support molecular-based phylogeny, viz. testing the hybrid origin hypothesis of C. sinensis from the perspective of palynology.

Plant Sampling and Identification
We collected 30 Chaenomeles species samples covering all currently accepted taxa of Chaenomeles, including five species and one interspecific hybrid cultivar. The collected living materials were photographed ( Figure 1) and identified on the basis of characters documented in Flora of China [2]. The botanical names were conformed from the International Plant Name Index (http://www.ipni.org (accessed on 22 March 2022)) and The Plant List (http://www.theplantlist.org/ (accessed on 22 March 2022)). Properly dried and mounted specimens were submitted to the Herbarium of Henan Agricultural University (HEAC) and the voucher numbers, along with the collection area and collector, are given in Table S1. Pollen grains from fresh mature but not dehiscent anthers were taken with sterilized forceps.
Diversity 2023, 15, x FOR PEER REVIEW 3 Therefore, the present study was designed with the aims (1) to document and i trate the morphology of optimally preserved pollen materials of the taxonomically p lematic genus Chaenomeles using both light and scanning electron microscopy; (2) to i tify informative (qualitative and quantitative) pollen features of taxonomical significa (3) to categorize Chaenomeles taxa based on statistical analysis (PCA and cluster analy and evaluate whether the results support molecular-based phylogeny, viz. testing th brid origin hypothesis of C. sinensis from the perspective of palynology.

Plant Sampling and Identification
We collected 30 Chaenomeles species samples covering all currently accepted ta Chaenomeles, including five species and one interspecific hybrid cultivar. The collected ing materials were photographed ( Figure 1) and identified on the basis of characters umented in Flora of China [2]. The botanical names were conformed from the Internati Plant Name Index (http://www.ipni.org (accessed on 22 March 2022)) and The Plant (http://www.theplantlist.org/ (accessed on 22 March 2022)). Properly dried and mou specimens were submitted to the Herbarium of Henan Agricultural University (HE and the voucher numbers, along with the collection area and collector, are given in T S1. Pollen grains from fresh mature but not dehiscent anthers were taken with steri forceps.

Observations Using LM
For light microscopy studies, pollens were treated by standard acetolysis accor to Erdtman's method [38]. The pollens collected from each sample were washed and immersed in glacial acetic acid. Glycerin jelly slides were prepared by a common p dure proposed by Bryant, Jones and Mildenhall (1990) [39], using glycerin-based sinensis (e,f), C. speciosa (g,h), C. ×superba (i,j) and C. thibetica (k,l).

Observations Using LM
For light microscopy studies, pollens were treated by standard acetolysis according to Erdtman's method [38]. The pollens collected from each sample were washed and then immersed in glacial acetic acid. Glycerin jelly slides were prepared by a common procedure proposed by Bryant, Jones and Mildenhall (1990) [39], using glycerin-based gel (stained with 1% safranin) from the sediment that contained pollen. We observed the slides using a light microscope (LM; BX41 Laboratory Microscope, Olympus, Melville, USA) at magnification ×1000. Six quantitative pollen characteristics (polar axis (P), equatorial diameter (E), colpus width (CW), colpus length (CL), endoaperture diameter (ED) and exine thickness (ET)) were measured using a digital camera for microscopes (MDX-30, Shinwoo Optics, Anyang, Korea) based on 25 pollen grains (n = 25) per sample.

Observations Using SEM
For SEM observations, we employed the critical point drying method (CPD) [40] to prepare pollen grains. The collected fresh anthers were fixed in 2.5% glutaraldehyde at room temperature, then post-fixed with 1% osmium tetroxide (0.1 M) at 4 • C for 2 h. All samples were subsequently dehydrated through an ethanol series, and finally rinsed with isoamyl acetate for 15 min. After dehydration, the anthers were critical-point dried and mounted on aluminum stubs using double-adhesive carbon tape. The pollen grains within the open locules of anthers were carefully expelled with dissecting needles. Using an ion-sputtering device (MC1000, Hitachi, Tokyo, Japan), the stubs were sputter-coated with gold for 180 s. The digital images of pollen grains at different magnifications were taken using SEM (S-4700, Hitachi, Tokyo, Japan), and the accelerating voltage ranged from 3-10 kV. The size measurements of quantitative pollen characteristics, including the equatorial diameter in polar view (EDPV), the widths of ridges (WRs) and valleys (WVs) of sexine ornamentations, the distance between the apices of two colpi (d), the endoaperture diameter (ED), the number of ridges per 25 µm 2 at the equatorial region of the exine (NR) and perforation diameter (PD), on SEM images of twenty-five pollen grains per sample, were made using Smile View software (ver. 2.1, JEOL Ltd., Tokyo, Japan). Polar area indices (PAI) were calculated as the ratio of d to EDPV. The descriptive terminology of ornamentation was mainly used by following previous studies [41][42][43][44].

Data Analysis
Palynological characters of the 30 individuals belonging to six taxa of Chaenomeles, comprising 13 quantitative characters (P, E, EDPV, P/E, PAI, CL, CW, ED, ET, WV, WR, NR, PD) and 2 qualitative characters (exine sculpture type and amb) were chosen for statistical analysis. Cluster analysis, based on a standardized pollen character data matrix and utilizing an unweighted pair-group method with arithmetic mean (UPGMA) [45], was performed to group Chaenomeles taxa and individuals into clusters on the basis of overall pollen character similarity. Palynological data of the outgroup species (Cydonia oblonga Mill.) were obtained from Radović et al. (2017) [46]. Principal component analysis (PCA) was employed to evaluate if pollen characters allowed taxa to be grouped, and to identify the characters that contribute most to the pollen morphological variability. We used untransformed and centered data to produce a covariance matrix and extract three eigenvectors from the matrix. We plotted the eigenvalues in a two-dimensional plot along the first two PCs. All computations were performed using the hclust function implemented in R (ver. 4.0.2, R Development Core Team 2020) for cluster analysis and the R vegan package for PCA. The relationships of quantitative pollen characteristics were determined by Pearson's correlation coefficients (Sokal and Rohlf, 1995) [47] and their variations among taxa were summarized by boxplot analysis. For individual pollen grain character, we utilized one-factor analysis of variance (ANOVA) to assess differences among all species studied. When critical differences were observed, multiple comparisons were applied with Tukey's test. The pollen terminology mostly followed that of Hebda and Chinnappa (1994) [28], Punt et al. (2007) [48], and we used Song et al. (2017) [31] to define the polar area index and the range of pollen size.

Results
The pollen morphology of all extant Chaenomeles species is described in terms of pollen size, shape, amb, colpus width, colpus length, polar area index, exine thickness, width of valleys, width of ridges, number of ridges per 25 µm 2 at the equatorial region of the exine, diameter of perforation and sexine ornamentation as follows. Pollen grains of Chaenomeles  Figures 2 and 3, respectively, and the pollen characteristics are summarized in Tables 1 and 2. Diversity 2023, 15, x FOR PEER REVIEW 5 of 23 The pollen morphology of all extant Chaenomeles species is described in terms of pollen size, shape, amb, colpus width, colpus length, polar area index, exine thickness, width of valleys, width of ridges, number of ridges per 25 µm 2 at the equatorial region of the exine, diameter of perforation and sexine ornamentation as follows. Pollen grains of Chaenomeles samples observed under LM and SEM are illustrated in Figures 2 and 3, respectively, and the pollen characteristics are summarized in Tables 1 and 2.

Shape and Size
Chaenomeles has isopolar, radially symmetrical and medium-sized pollen grains (Table 1). With respect to polar diameter, the largest value was observed in C. thibetica (P = 40.61 ± 1.48 µm), and the smallest in C. sinensis (P = 33.98 ± 0.83 µm). For equatorial diameter, the variation was less evident, with the maximum value observed in C. japonica (E = 34.92 ± 1.04 µm) and the minimum value examined in C. speciosa (E = 32.11 ± 0.93 µm). The shape of the pollen grains, obtained through the ratio of polar diameter and equatorial diameter (P/E), varies from prolatespheroidal to subprolate (P/E = 0.98-1.35; Figure 2). The outline in the polar view (amb) is mostly subcircular and subangular in C. japonica and C. speciosa, subcircular grains were mainly observed in C. cathayensis, circular and subangular grains were mainly observed in C. sinensis, and subangular grains were mainly observed in C. ×superba and C. thibetica ( Table 1). The variation in shape and size is wider in ACE pollen grains (P = 32.78-42.74 µm, P/E = 0.98-1.35) than in CPD pollen grains (P = 16.36-25.93 µm, P/E = 0.66-0.87). According to the results of the ANOVA, significant differences were observed in pollen diameter and P/E ratio at the species level ( Figure 2). The frequency of deformed pollens varied among taxa and ranged from 1.65% (C. cathayensis) to 17.50% (C. ×superba) in non-hybrids. In hybrid C. ×superba, the observed percentage of deformed pollen (14 grains in total) was higher than in the parental species.

Polar Area and Apertures
The polar area is generally small (PAI = 0. 23-0.35) in Chaenomeles species, with that of C. sinensis being the largest and significantly larger than that of the other species, and that of C. thibetica being the smallest (Figure 3p-r; Table 2). All taxa produce tri-colporate pollen grains in monads (Figures 2 and 3). Simple colpi were arranged meridionally and symmetrically, with acute and pointed ends. The colpi are slightly invaginate (shallow), but distinct in the LM equatorial view. Their lengths (CL) are long and fairly varied, from 28.44 µm (C. sinensis; Figure 2c) to 34.30 µm (C. thibetica; Figure 2i), and constitute, on average, 78.02-88.75% of the polar diameter. The colpi length exhibit strong correlation with the polar axis length (r = 0.84, p < 0.001). Colpi are fusiform in outline and their width is diverse and mostly widest in the equatorial region. The significantly wider colpi (equatorial extent) in Chaenomeles occurred in C. sinensis (8.53 µm, Figure 2f) and C. ×superba (7.69 µm, Figure 2k), while the narrowest ones were found in C. speciosa (7.06 µm, Figure 2j). All species have fusiform colpus opercula. The opercula are wide and flattened to faintly convex at the equator, with corrugated surfaces (Figure 2). Endoapertures were observed in all studied taxa. They are circular or elliptic in outline, and usually located in the middle of colpi, with diameters ranging from 5.71 to 8.80 µm (Figures 2 and 3; Table 1).

Exine Ornamentation
All Chaenomeles taxa have a sexine ornamentation pattern consisting of supratectal ridges and valleys, with perforations of varying sizes occurring in the valleys (Figure 3). A distinct zone of densely packed ridges, regularly aligned and parallel to the colpus, occurs at the margins of the colpus. The number of ridges in 25 µm 2 ranges between 22 and 54. Two types (I, II) and two subtypes (II-A, II-B) of sexine ornamentation can be recognized based on the elongation patterns of the striae and the diameter of perforations (PR) (Tables 1 and 2, Figure 3). The exine thickness varies from 0.58 to 1.29 µm (Table 1; Figure 2). Type I striate with microperforations: Ridges are long, continuously elongated with few anastomoses, and nearly all run parallel to colpi and continue over the poles. The extension of striae before ending or changing direction are usually more than 10 µm. This type contains tectal microperforations similar to pinpricks (0.03-0.06 µm in diameter), which were often obscured by ridges but clearly visible at poles. Type I ornamentation was exclusively observed in C. sinensis (Table 1; Figure 3g-i).
Type II striate with macroperforations: this type is subdivided according to the length of ridges.
Subtype II-A (twisted striate): ridges are medium to long in length, 0.14-0.22 µm wide, and mostly oriented parallel to the colpus and pole axis, but regularly, they also form some curving and looping (fingerprint-like twists) at the inter-colpium region. This type includes pollen grains with wider valleys as well as larger perforations (0.12-0.30 µm in diameter, Figure 2). Taxa with subtype II-A ornamentation are: C. cathayensis (Figure 3a-c), C. japonica (Figure 3d-f), C. thibetica (Figure 3p-r).
Subtype II-B (reticulate striate): ridges are short, 0.13-0.23 µm wide, forming a complex interwoven pattern. The extensions of ridges before an end or a change in direction are usually less than 4 µm. The perforations at the bottoms of valleys are 0.12-0.20 µm in diameter. Taxa with subtype II-B ornamentation are: C. speciosa (Figure 3j-l), C. ×superba (Figure 3m-o).

Multivariate Analysis
Statistically significant differences were observed among the Chaenomeles taxon based on the quantitative pollen grain features analyzed here (p < 0.05). The statistical analyses are illustrated and summarized in Figure 4. The PCA graph ( Figure 5) presented the projections of the pollen characteristics in a multi-dimensional space and the first and second components accounted for 68.8% of the total variance. PC1 explains 49.9% of the overall variance and is related to the polar area index, number of ridges per 25 µm 2 , diameter of perforation, amb and colpus length, whereas PC2, which accounted for 18.9% of the variance, is associated with exine sculpture type, equatorial diameter in polar view, polar axis and colpus width (Table 3). Overall, the number of ridges per 25 µm 2 at the equatorial region of the exine and equatorial diameter were the most distinctive and valuable characters, since they accounted for the largest amount of relative variation. C. sinensis, which has the unique sexine ornamentation type I, is positioned on the negative side of the PC 1 axis and is more separated from the other species in the PCA plot. The taxa of sexine ornamentation type II-B, including C. ×superba and C. speciose, are grouped on the negative side of the PC 2 axis. C. thibetica and C. cathayensis are in close similarity with regard to the data from P, P/E, CL, WV, PR and are distributed on the positive side of the PC 2 axis ( Figure 5). R PEER REVIEW Figure 4. Boxplot and dot-plot graphs for the quantitative pollen characters in the studied Chaenomeles taxa. Different superscript letters d significant differences between species (p < 0.05). The black dots located outside the whiskers of the box plot are considered outliers. Figure 4. Boxplot and dot-plot graphs for the quantitative pollen characters in the studied Chaenomeles taxa. Different superscript letters demonstrate statistically significant differences between species (p < 0.05). The black dots located outside the whiskers of the box plot are considered outliers.
The cluster analysis revealed that the palynological data obviously divides all of the taxa and individuals studied into three groups, concurrent with the PCA. The group comprising C. sinensis was firstly set apart from the other taxa, due to its significantly lower values of P/E ratio, colpus length, width of ridges and perforation diameter, and its larger polar area (Figures 3 and 6). The taxa with exine sculpture subtype II-A, comprising samples of C. thibetica, C. cathayensis, C. japonica, formed a second group. Within this group, C. thibetica and C. cathayensis are in close similarity as to pollen and colpus dimensions and are distributed on the positive side of PC 1 (Figures 5 and 6). The rest of the taxa (C. ×superba, C. speciosa) sharing exine sculpture subtype II-B and similar amb are clustered in the third group (Figures 5 and 6). The samples from the same species tend to be clustered together, except for C. superba-AD5, which grouped with two samples of C. speciosa.    ples of C. thibetica, C. cathayensis, C. japonica, formed a second group. Within this group, C. thibetica and C. cathayensis are in close similarity as to pollen and colpus dimensions and are distributed on the positive side of PC 1 (Figures 5 and 6). The rest of the taxa (C. ×superba, C. speciosa) sharing exine sculpture subtype II-B and similar amb are clustered in the third group (Figures 5 and 6). The samples from the same species tend to be clustered together, except for C. superba-AD5, which grouped with two samples of C. speciosa.

Discussion
In this study, LM and SEM were applied to observe the pollen morphological features of extant species of Chaenomeles in China. Statistical analyses of the morphological variability were undertaken for the first time to identify qualitative and quantitative characteristics that were valuable and informative to distinguish Chaenomeles species.
Pollen grains of Chaenomeles are medium in size with a prolatespheroidal to subprolate shape and are tri-colporate, in accord with the overall pollen features of Rosaceae [27,28,30,31,34]. However, we discovered a statistically significant variance in pollen size and shape between species (Figures 2 and 3). Chaenomeles species also differ with regard to polar surface area and colpus size (Figure 3). These results indicated that pollen and colpus dimensions, which are controlled to a large extent by genes, are suitable characters for species characterization. The only previous palynological study of Chaenomeles simply followed an air-drying method and reported a range of pollen sizes (P: 33.24-48.27µm; E: 18.93-25.58µm), differing to a large extent from those we obtained here [37]. In fact, preparation treatments can affect pollen size and shape [31,40,[49][50][51][52]. In comparison with the previous SEM images of Zang and Ma (2004) [37], we believe that the pollen grains of Chaenomeles were highly fragile and showed pressure distortion (pollen wall folding back along the apertures) after direct air drying, which possibly led to biases in measurement. Therefore, suitable artificial drying methods should be used to fix and further observe the natural pollen state of this genus. In this study, ACE and CPD methods were both employed for pollen preparation, and the latter is superior in maintaining the natural pollen size and shape, leaving the apertural membranes undestroyed [53]. It is suggested here that CPD pollen is always smaller than ACE pollen (mean by 49.12%; range: 28.48-68.44%; Table 2) like other palynological studies have found [30,40,54]. In general, the pollen size of Chaenomeles was larger than its most closely related genera, as reported earlier, including Cydonia [46], Photinia [55], Sorbus [56] and Dichotomanthes Kurz [57].
Previous palynological studies demonstrated that sexine ornamentation features are highly variable and informative for Rosaceae pollen at the genus as well at the species level [27,28,[33][34][35][36][37]58,59]. The most important features include the number, interval, width, length and direction of striae, as well as the diameter and density of perforations [27,30,31,34,36]. Our results do not support the diagnostic significance of the number, interval and width of striae, since except for C. sinensis, other taxa from Chaenomeles were in close similarity based on measurements of NR, WV and WR, whereas features such as the length or direction of striae and diameter of perforation, as a result of their considerable variations, were identified to be of taxonomic value and informative in species delimitation (Table 2, Figure 3). Similar results were obtained from the pollen study on different genera (Spiraea L., Sorbaria (Ser.) A. Braun, Rosa L., etc.) within the family Rosaceae [30,31,34]. Hebda and Chinnappa (1994) [28] classified the pollen sculpturing type of Rosaceae into six major categories: striate macroperforate, striate microperforate, tuberculate, microverrucate, verrucate and perforate without supratectal features. They included Chaenomeles in type I (the third subtype), which were described as having short ridges (waving and crossing) separated by valleys with larger perforations in the 0.1-0.2 µm diameter range. Nevertheless, there is mounting evidence that even in this small genus, the inclusion of taxa into one type is too general. According to Zang and Ma (2014) [37], the above-mentioned type I was only quoted for C. japonica, and they newly distinguished two types of exine sculpture, one characterized by fine stripes nearly parallel to each other (e.g., C. sinensis), and the other with irregular striae and rare or no perforations (e.g., C. thibetica). Such division is not adopted here, firstly because it was based on air-dried pollens, which in comparison with the CPD pollens we obtained here ( Figure 3) were subjected to obvious distortion and shrinkage (e.g., E decreased by 14.51-52.09%), leaving apertures unrecognized and perforations obscured, and secondly because it is descriptive, relying neither on measurements nor statistical analysis of exine sculpture characters, and is thus difficult to compare across studies. Notwithstanding, our study agreed with Zang and Ma (2014) [37] in that species from Chaenomeles also belong to sculpture types other than type I as well as to types that are not described by Hebda and Chinnappa (1994) [28].
In the current study, we distinguished two types and two subtypes of sculpture in Chaenomeles based on ridge patterns and perforation size (macroperforation/microperforation). While a majority of the members in Chaenomeles possess striate type II, C. sinensis is clearly distinct in having microperforations and conspicuous long ridges parallel to the colpus (type I; Figure 3g-i). In addition, C. sinensis can be separated as a basal group on the basis of significant differences in the qualitative pollen features we analyzed here (Figures 5 and 6). In fact, C. sinensis was previously treated as Pseudocydonia, a monotypic genus closely related to Cydonia. The taxonomical position of Pseudocydonia has been the subject of much debate [21,60]. Previous phylogenetic studies revealed strong conflicts between chloroplast and nuclear data in the placement of C. sinensis [3,12,13,60,61], indicating that this species might originate via hybridization between members of the ancestral lineages of Cydonia and Chaenomeles. Pollen evidence from our study supports the hybrid origin hypothesis, as the pollen type of C. sinensis is intermediate between that of typical Chaenomeles (sharing type II) and Cydonia in that it has conspicuous striate resembling Chaenomeles, and longitudinal fine ridges and tectal microperforations resembling Cydonia [37,46]. Furthermore, C. sinensis also shows some morphological resemblance to both of the putative parental taxa, with the basally fused styles and 25 or fewer stamens of Cydonia and the serrate leaves, glabrous fruits, and deciduous sepals of Chaenomeles [2,21,22]. Since C. sinensis is morphologically anomalous in either Cydonia or Chaenomeles, this study supports the placement of this species in the monotypic genus Pseudocydonia.
As for the rest of the taxa, C. cathayensis, C. japonica and C. thibetica belonged to the same exine ornamentation subtype (type II-A; Figure 3a-f,p-r); yet they were discriminated as separate species based on the significant differences in perforation diameter (PR) and ridge width (WR) (Figure 4). The statistical analysis supported the closer relationship between C. cathayensis and C. thibetica, since they were more similar in pollen shape (P/E ratio), colpus length (CL) and colpus width (CW) (Figure 2). This corroborates the results of Lo and Donoghue (2012) [13] and Sun et al. (2020) [3], which, based on molecular phylogenetic analysis, inferred C. thibetica and C. cathayensis as sister species with high bootstrap support. In this study, pollen grains of the hybrid taxa C. ×superba and parents (C. japonica × C. speciosa) were examined in detail for the first time using both LM and SEM. The majority of diagnostic pollen features studied (P/E, CL, CW, PR) for C. ×superba are intermediate between that of C. japonica and C. speciosa (Tables 1 and 2; Figure 2). However, the exine pattern of C. ×superba was classified as striate-reticulate with macroperforations, identical to that of C. speciosa (subtype II-B; Figure 3m-o, j-l). In addition, C. ×superba have comparably thinner pollen walls than that of other Chaenomeles species, including the parents. When observing SEM images, we noticed a higher proportion of deformed pollen grains for C. ×superba (17.50% abnormal pollen grains) than for the remaining taxa. The pollen deformation included mainly the loss of turgor pressure and changes in pollen shape. This feature was frequently found in ornamental plants and hybrids, and has been reported in other hybrid members of related genera (Malus, Crataegus L., etc.) in Rosaceae [36,44,62].

Conclusions
Chaenomeles pollen grains are medium sized, prolatespheroidal to subprolate and tricolporate, with striate sexine ornamentation. The pollen features of size, polar area, colpus dimension and sexine sculpture were verified by statistical analysis to be useful criteria for delimiting species. Two sexine sculpture types (type I-striate with microperforations, type II-striate with macroperforations) and two subtypes (twisted-striate, reticulate-striate) were recognized in the species studied depending on direction and length of the ridge patterns, which might be of systematic importance. C. sinensis is clearly distinct from the other Chaenomeles species on the basis of the unique sexine ornamentation type I, as well as the significant differences in qualitative pollen features we analyzed, which supports the placement of this species in the monotypic genus Pseudocydonia. This study is the first comprehensive investigation of the pollen morphology of Chaenomeles species based on LM and SEM and demonstrated the diagnostic significance of pollen features in the taxonomy of closely related species. Furthermore, the present data is beneficial to a comprehensive understanding of pollen morphology in Chaenomeles as part of ongoing palynological studies of Rosaceae that strive to provide an overview of the family's pollen micromorphology and character evolution.