Pollen morphology of Polish species from the genus Rubus L. (Rosaceae) and its systematic importance

The genus Rubus L. (Rosaceae) not been investigated satisfactorily in terms of palynology. This genus is taxonomically very difficult due to the large number of species and problems with their delimitation, as well as very different distribution areas of particular species. The aim of this study was to investigate pollen morphology and for the first time the ranges of intrageneric and interspecific variability of Rubus species, as well as verify the taxonomic usefulness of these traits in distinguishing studied taxa from this genus. The selected species of the genus Rubus were analysed for 11 quantitative pollen characteristics and the following qualitative ones: exine ornamentation, pollen outline and shape, as well as bridge structure. Analyses were conducted on a total of 1740 pollen grains, which represent 58 blackberry species belonging to a majority of subgenera and all the sections and series found in Poland. The most important characters included exine ornamentation (exine ornamentation type, width and direction of grooves and striae, number and diameter of perforations) and length of the polar axis (P). The arrangement of the examined species on the dendrogram does not corroborate division of the genus Rubus into subgenera, sections and series currently adopted in taxonomy. This fact is not surprising because the taxonomy of the genus was not based on pollen characters. Pollen features should be treated in taxonomy as auxiliary, because they fail to differentiate several (10) individual species, while the other ones create groups with similar pollen traits.


Introduction
Rubus L. is a large and diverse genus in the Rosaceae family with a worldwide distribution, including hundreds or even thousand of published species names and infrageneric taxa [1,2]. Depending on which classification you follow, historic or modern, the number of Rubus species may vary from 429 to 750 or up to 1000 worldwide [3][4][5][6][7][8][9].

Pollen morphology
The collected plant material was stored in the herbarium of the Faculty of Forest Botany of the Poznań University of Life Sciences (PZNF), which did not require any permits to conduct research.
The study was conducted on 58 Polish and European Rubus species, which represent four out of five subgenera, all three sections and all 23 series of blackberries found in Poland, including all six Polish endemic species (R. capitulatus, R. chaerophylloides, R. ostroviensis, R. posnaniensis, R. seebergensis and R. spribillei). A list of the species analysed with their affiliation to particular taxa is shown in Table 1.
In this paper, the taxonomic classification of the studied taxa from the genus Rubus was adopted from Zieliński [16], with further modifications [18]. The verification of the taxa was made by Prof. Jerzy Zieliński (Institute of Dendrology, Polish Academy of Sciences in Kórnik), a batologist-taxonomist specialising in the genus Rubus.
Several, randomly selected inflorescences (flowers) were collected from 58 natural blackberry localities in Poland (Table 2).
Pollen grains were acetolysed according to the method of Erdtman [49]. The inflorescences collected from the herbarium were placed in tubes and then centrifuged with glacial acetic acid. Grains were mixed with the acetolysis solution, which consisted of nine parts acetic anhydrite and one part concentrated sulphuric acid. The mixture was then heated to boiling and kept in the water bath for 2-3 min. Samples were centrifuged in the acetolysis mixture, washed with acetic acid and centrifuged again. The pollen grain samples were then mixed with 96% alcohol and centrifuged 4 times, with processed grains subsequently divided into two groups. One half of the processed sample was immersed in an alcohol-based solution of glycerin for LM, while the other was placed in 96% ethyl alcohol in preparation for scanning electron microscopy (SEM). The SEM observations were made using a Zeiss Evo 40 and the LM measurements of acetolysed pollen grain were taken using a Biolar 2308 microscope at a magnification of 640x. Pollen grains were immersed in glycerin jelly and measured using an ocular eyepiece with a scale. Measurements taken from 30 mature, randomly selected, properly developed pollen grains were made by using the light microscopy (LM), with 1740 pollen grains measured in total. Measurement results were then converted into micrometres by multiplying each measurement by two.
Exine ornamentation types (I-VI) were identified based on the classification proposed by Ueda [47]. The types and subtypes of the striate exine ornamentation were characterised by the height and width of grooves, width of striae and the number and diameter of perforations.

Statistical analysis
The normality of the distributions for the studied traits (P, E, Le, d, Exp, Exe, P/E, Le/P, d/E, Exp/P and Exe/E) was tested using Shapiro-Wilk's normality test [53]. Multivariate analysis of  variance (MANOVA) was performed on the basis of the following model using the MANOVA procedure in GenStat (18th edition): Y = XT+E, where: Y is the (n×p)-dimensional matrix of observations, n is the number of all observations, p is the number of traits (in this study p = 11), X is the (n×k)-dimensional matrix of design, k is the number of species (in this study k = 58), T is the (k×p)-dimensional matrix of unknown effects and E-is the (n×p)-dimensional matrix of residuals. Next, one-way analyses of variance (ANOVA) were carried out to determine the effects of species on the variability of examined traits, for each trait independently, on the basis of the following model: y ij = μ+τ i +ε ij , where: y ij is the jth observation of the ith species, μ is the grand mean, τ i is the effect of the ith species and ε ij is an error observation.
The arithmetical means and standard deviations of traits were calculated. Moreover, Fisher's least significant differences (LSDs) were also estimated at the significance level α = 0.001. The relationships between observed traits were assessed on the basis of Pearson's correlation. Results were also analysed using multivariate methods. The canonical variate analysis was applied in order to present multitrait assessment of similarity for the tested species in a lower number of dimensions with the least possible loss of information [54]. This makes it possible to illustrate variation in species in terms of all the observed traits in the graphic form. The Mahalanobis distance was suggested as a measure of "polytrait" species similarity [55], which significance was verified by means of critical value D α called "the least significant distance" [56]. Mahalanobis distances were calculated for species. The differences between the analysed species were verified by cluster analysis using the nearest neighbour method and Euclidean distances [57]. All the analyses were conducted using the GenStat (18th edition) statistical software package [58].

General morphological description of pollen
A description of pollen grain morphology of the Rubus species studied is given below and illustrated with several SEM photographs (Figs 1-3). The morphological observations for the other quantitative characters of pollen grains are summarised in Table 3.

PLOS ONE
Pollen morphology of Polish species from the genus Rubus L.

PLOS ONE
Pollen morphology of Polish species from the genus Rubus L.  Table 3).
pollen grains belong to the upper limit of small pollen or to the lower medium-sized pollen range.
The outline in the polar view was mostly circular with obtuse apices, more rarely elliptic, whereas in the equatorial view the outline was mostly elliptic, rarely circular (Fig 1).
The mean P/E ratio was 1.19, ranging from 0.85 in R. pedemontanus to 1.71 in R. saxatilis (Table 3). On average the P/E ratio values were always above 1 and they ranged from 1.05 in R. pedemontanus to 1.32 in R. chaerophylloides. Pollen grains of the species examined were most frequently subprolate (57.3% -997 pollen grains) or prolate-spheroidal (24.3% -422), rarely      Table 3). The relative thickness of the exine (Exp/ P ratio) averaged 0.07 (0.02-0.18) and (Exe/E ratio) 0.08 (0.02-0.14). The above results were similar, indicating a more or less equal exine thickness along the entire pollen grain (Table 3).
In all the studied species, exine ornamentation was striate-perforate and very rarely striate, with the exception of R. odoratus, which had a striate-verrucate ornamentation with small perforations (Fig 3). Exine ornamentation elements were highly variable (Fig 3). Striae and grooves usually ran parallel to colpori and the polar axis, but frequently they also formed fingerprint-like twists. Striae were straight or forked and of varying length, width and height.
The investigated pollen of the individual Rubus species was classified according to the striate exine ornamentation classification proposed by Ueda [47] into four types (I-III and V) and five subtypes (I A, II A,B and III A,B). The cited author distinguished six types (I-VI) and six subtypes (I-III, each A and B). In our study types IV, VI and subtype IB were not found (Fig 3,  Table 4). The greatest number of species (18) belonged to the IIA subtype, which was characterised by fairly distinct striae, narrow grooves and frequently by prominent, numerous perforations. Subtypes IA, IIA/IIB, IIB and IIIA were represented by a relatively large number of species (8, 11, 8 and 9 species, respectively), while types IA/IIA, IIIB and V-by only one species. Among the 58 examined species, 12 had two types of exine ornamentation (Fig 3, Table 4).
In most of the species (56 of the 58), elliptic or circular perforations of different diameters (0.05-0.4 μm) were found at the bottom of the grooves (Fig 3). The perforations were not found in R. canadensis and R. czarnunensis. In the majority of the species studied the perforations were small, with similar diameters (0.1-0.2 μm) and more or less numerous, with the exception of R. bifrons, R  Pollen grains usually had three apertures-colpori. Ectoapertures-colpi were arranged meridionally, regularly, they were more or less evenly spaced and long, at a mean length of 21.23 (14-32) μm (Table 3; Fig 4D-4F). On average, the length of colpi constituted 83% (from 60 to 100%) of the polar axis length, with the shortest colpi found in R. xanthocarpus (16.2 μm) and the longest in R. corylifolius (25.3 μm). Colpi were fusiform in outline. Their width was variable and usually greatest in the equatorial region. Sculpturing of ectocolpus membranes approached rugulate, rarely partly psilate (Fig 4D-4F). Colpus margins frequently had small undulations (Fig 4D-4F).
In all of the species studied the colpus was crossed at the equator by a bridge dividing it into two parts, formed by two bulges of the ectexine that meet in the middle (Fig 4A-4C). The bulges were of the same or unequal length.
Endoapertures were usually located in the middle of colpi, less frequently asymmetrically, usually singly and very rarely in pairs. They were circular or elliptic in outline with irregular margins (Fig 4D-4F).

Pollen key
Pollen key can be seen as a summary of the outcome of our study thus it has been placed at the very end of this chapter.  15.29)] showed variability of the tested species at a significance level α = 0.001. The mean values and standard deviations for the observed traits indicated a high variability among the tested species, for which significant differences were found in terms of all the analysed morphological traits ( Table 3).
The correlation analysis indicated statistically significant correlation coefficients for 25 out of 55 coefficients (Table 5). A total of 16 out of 25 significantly correlated pairs of traits were characterised by positive correlation coefficients. In the case of 30 pairs of traits, no significant correlation was established.
In the presented dendrogram, as a result of agglomeration grouping using the Euclidean distance method, all the examined Rubus species were divided into four groups (Fig 5). The first group (I) comprised one species-R. czarnunensis, while the second one (II) four species (R. dollnensis, R. corylifolius, R. chaerophylloides and R. phuhianus). The third group was divided into two subgroups: III A-R. camptostachys, R. xanthocarpus, R. clussi, R. odoratus, and III B-including all the other species from this group. The fourth group (IV) comprised R. canadensis, R. capitulatus, R. acanthoides and R. spribillei.
Individual traits were of varying importance and had different shares in the joint multivariate variation. A study on the multivariate variation for species includes also identification of the most important traits in the multivariate variation of species. Analysis of canonical variables is a statistical tool making it possible to solve the problem of multivariate relationships. Fig 6 shows the variability of the pollen grain features in 58 studied Rubus species in terms of the first two canonical variables. In the graph the coordinates of the point for particular shrubs were the values for the first and second canonical variable, respectively. The first two canonical variables accounted for 56.75% of the total multivariate variability between the individual species. Five groups of species were distinguished (Fig 5). A majority of the examined species were found in the first group (I), which means that they had more or less similar pollen features. Only one up to maximum three species (II-R. capitulatus, III-R. xantocarpus, IV-R. acanthoides and R. spribillei, and V-R. corylifolius, R. dollnensis, and R. czarnunensis) fell into the other four groups (Fig 6). Pollen grains of R. capitulatus were the most different from those of the other species (large, with a thin exine and the P/E ratio usually prolate-spheroidal).

PLOS ONE
Species from groups IV and V had the largest pollen grains and R. xantocarpus (group III)the smallest ones. The most significant, positive, linear relationship between the first canonical variables was found for P, E, Le and d, while it was negative for Exp/P and Exe/E ( Table 6). The second canonical variable was significantly negatively correlated with Exp, Exe, Exp/P and Exe/E ( Table 6). The greatest variation in terms of all the traits jointly (measured Mahalanobis distances) was found for R. canadensis and R. capitulates (the Mahalanobis distance between them amounted to 8.24). The greatest similarity was found for R. lamprocaulos and R. hevellicus (0.313).

Discussion
Similarly to a majority of palynologists, the authors of this study maintain that exine ornamentation features were diagnostic, that means they allow for differentiate species within the genus Rubus [24, 25, 27-31, 33, 34, 38, 39, 42, 46, 59]. The most important exine ornamentation traits include the width, number and course of grooves (muri) and the width of the striae as well as the number and diameter of perforations [31,33,34,42,46,[59][60][61]. Some authors considered pollen size and shape as potentially important features in the diagnosis of the analysed Rubus species [27,28,33], while others claim that they have no diagnostic significance [31,45,46]. Based on our results, we partially agree with the opinion of these former, because the length of the polar axis (P) has been an important feature.
Ueda & Tomita [61] and Ueda [47] distinguished six types and six subtypes of exine ornamentation in species and other taxa from the genus Rosa and the family Rosaceae, including the genus Rubus. In the current study they were classified into four types (types IV and VI were not identified) and five subtypes (I A, II A, B, III A, B). Our results were similar to the cited authors, since most of the examined pollen belonged to the IIA and IIIA subtypes and no grains were found in the very rarely represented types IV and VI or subtype IB. The only species described both by Ueda [47] and in our study was R. odoratus. Ueda [47] described it as a type VI and we as type V.
The research results obtained in this study confirmed the diagnostic significance of the number and diameter of perforations, found by Hebda & Chinnappa [38,39], Monasterio-Huelin & Pardo [28], Tomlik-Wyremblewska [31], Li et al. [42], Wrońska-Pilarek et al. [33] or Ghosh & Saha [59], because these traits allowed to distinguish certain Rubus species (see: pollen key). On the other hand, groups of species from different sections possess similar numbers of perforations (e.g. R. opacus from the series Rubus, R. canadensis from the series Canadenses or R. henrici-egonis from the series Discolores). However, also species from many different Many studies reported that the bridges are located in the most of studied Rubus species. [28,31,33,46]. They were wide, well-developed and with margins. In blackberries Tomlik-Wyremblewska [31] distinguished two bridge types, with margins stretched or constricted at the equator. In our study, bridges were observed in all the analysed blackberry species and this structure was not used as a basis for the identification of species, because its characteristics were too similar. Besides, it usually appeared in mature pollen grains, so it could not be noticed when analysing pollen at other developmental stages.
The arrangement of the investigated species on the dendrogram (Fig 5) does not corroborate the division of the genus Rubus into subgenera, sections and series [16], currently adopted in taxonomy.Species from three different subgenera (R. saxatilis and R. xanthocarpus from the subgenus Cylactis, R. odoratus from the subgenus Anoplobatus and R. idaeus from the subgenus Idaeobatus) were found in the same group III, with most of the species from a large subgenus Rubus. Similar results were obtained for the three sections from the subgenus Rubus (Rubus, Corylifolii and Caesii). Thus, R. caesius from the section Caesii and R. gothicus, R. camptostachys, R. mollis or R. fabrimontanus from the section Corylifolii were found in group III, with the species representing the most numerous third section of Rubus. Also in the case of the series it were not observed that species belonging to these taxa formed separate groups (Figs 5 and 6). Other genera of the family Rosaceae (e.g. Spiraea, Rosa, Crataegus) showed a correlation between pollen morphology and intrageneric taxonomic classification [62][63][64]. In Rubus the lack of dependence could be the result of apomixis, defined as the replacement of the normal sexual reproduction by asexual reproduction, without fertilisation, which could reduce natural variability.