Systematic importance of morphological features of pollen grains of species from Erica (Ericaceae) genus

The Erica genus has as yet not been investigated satisfactorily in terms of palynology. Its complicated taxonomic system, large number of species, as well as its extensive but disrupted range of occurrence, all contribute to the fact that few researchers have undertaken investigations of this species. It was assumed that the research results would be representative thanks to a complex comparative analysis of all diagnostical, morphological pollen features performed on properly selected plant material, representing the most important distinguished intrageneric taxons at the present time (45 species from all five subgenera and 22 sections), both discriminated pollen dispersal units (tetrads and monads) as well as the main centres of genus occurrence and diversification (species from Europe, the Republic of South Africa (RSA) and Madagascar). The study revealed that the diagnostic features of the pollen grains studied were: pollen dispersal unit, exine ornamentation, P/E ratio, tetrad diameter (D) and length of polar axis (P). On the basis of these traits, 14 Erica species (six creating monads and eight—tetrads) were distinguished which, in the case of pollen features, constitutes a significant number. Other heaths created small groups, usually containing two or three species, but up to seven species. The present study, based on the highest number of Erica species (45) analysed so far, corroborated the view that an examination of palynological features may assist in clarifying classification systems for the large and taxonomically very difficult Erica genus, in particular, at the level of the subgenus and section, but also at species level. The results obtained indicate the need to continue palynological investigations on the Erica genus.


Introduction
The family Ericaceae comprises eight subfamilies, approximately 125 genera and 4100 species [1]. Recently, a revised phylogenetic classification of the Ericaceae family was proposed by [2], which recognizes eight subfamilies and 20 tribes, and identifies Cyrilla (Cyrillaceae) as a sister to the Ericaceae. However, tribal and generic circumscriptions thus far remain ambiguous SWO  investigations based on such a large number of species have hitherto not been the object of palynological studies. Until now, the most extensive research comprised 23 Erica species [36], of which only five (E. arborea, E. axillaris, E. cinerea, E. nabea, E. tetralix) were also examined in the study presented. Previously, fossil pollen grains of 17 European Erica species was described [35], of which five (E. arborea, E. australis, E. cinerea, E. erigena, E. tetralix) also appeared in the present study. In addition, the interspecific variability of the pollen grains of the Erica taxa under investigation has not yet been comprehensively analysed. Another advantage of this work is the fact that it examined endemic species from Madagascar not investigated thus far and earlier included in the Philippia genus. This allowed a comparison of the pollen grains representing this centre of occurrence with species representing the so-called minor genera, as well as the heather species sensu stricto.

Palynological analysis
They were analysed 45 Erica species; 39 species with terads and six species with monads. A list of the analysed taxa, with their affiliation to particular subgenera and sections, is shown in Table 1.
In this paper, the taxonomic classification of the studied taxa from Erica genus was adopted from publication [10], with further changes [11,12,21,40]. The sections names for species from South Africa follows [9] and for species fom Europe follows [41]. The verification of the taxa from South Africa was made by Prof. Edward Oliver (Stellenbosch Herbarium, National Botanical Institute, South Africa), the taxa from Europe and Madagascar were reviewed by Prof. Piotr Szkudlarz (Adam Mickiewicz University in Poznań, Poland)-the outstanding taxonomists specializing in the Erica genus. In Table 1, heather species from Minor Genera, which were included in the Erica genus [6], were excluded from the system because, so far, they had not been allocated to any section. Similar approach was adopted in the case of species derived from Madagascar, which were earlier placed in the Philippia genus [21].
The plant material were collected by from 45 natural Erica sites in years: 2005 (South Africa-Republic of South Africa), 2010 (Spain) and 2014 (Madagascar) ( Table 1).
Several randomly selected inforesecences or flowers were collected from each individual plant. For each analysed species 30 mature, correctly formed pollen grains was measured by LM. In total, 1350 pollen grains were analysed.
All samples were acetolysed according to Erdtman's method [42], with insignificant modifications. Shrivelled, ground flowers or inflorescences were blent with acetolysis mixture, consisted of acetic anhydrite (9 portions) and concentrated sulphuric acid (1 portion). Then suspension was heated to boiling point and kept in water bath for 2-3 minutes. Next the samples were centrifuged in acetolysis mixture, washed with acetic acid and centrifuged again. Finally they were both mixed with alcohol 96% and centrifuged four times. Prepared material was divided into two parts. One half was immersed in alcohol solution of glycerine (for LM) and the other in ethyl alcohol 96% (for SEM).
The observations were carried out with light microscope (Biolar 2308) and scanning electron microscope (Philips SEM-515) on acetolysed pollen grains and the measurements were carried out with LM on acetolysed pollen grains to. Most of the measurements were done on the LM at magnification of x640. Pollen grains were prepared in glycerine jelly and measured using the eyepiece (ocular) with scale and then measurement results were recalculated into micrometers by multiplying them by 2.
The pollen grains were analysed for 11 quantitative features, i.e. tetrad diameter (D), length of polar axis (P) and equatorial diameter (d in tetrads or E in monads), length of ectoaperture (2f in tetrads or Le in monads), apocolpial (Ex) and septal (Se) exine thickness and for a monads P/E, Le/P, Ex/E ratios and for a tetrads D/d, P/d, 2f/D, Ex/Se and Ex/d ratios; and the following qualitative ones: outline, shape and exine ornamentation. Tetrads were measured in the distal face and monads in polar view.

Statistical analysis
Monads and tetrads were analysed independently. Firstly, the normality of the distributions of the studied traits (for monad: P, E, Le, Ex, P/E, Le/P and Ex/E; for tetrad: D, d = E, P, 2f = Le, Se, Ex, D/d, P/d, 2f/D and Ex/Se) were tested using Shapiro-Wilk's normality test [44]. Multivariate analysis of variance (MANOVA) was performed on the basis of following model using a procedure MANOVA in GenStat 17th edition: Y = XT+E, where: Y is (n×p)-dimensional matrix of observations, n is number of all observations, p is number of traits (in this study p = 7 for monad and p = 11 for tetrad), X is (n×k)-dimensional matrix of design, k is number of species (in this study k = 6 and k = 39 for monad and tetrad, respectively), T is (k×p)-dimensional matrix of unknown effects, E-is (n×p)-dimensional matrix of residuals. Nextly, oneway analyses of variance (ANOVA) were performed in order to verify the zero hypothesis on a lack of species effect in terms of values of observed traits, for each trait independent, on the basis of the following model: y ij = μ+τ i +ε ij , where: y ij is jth observation of ith species, μ is grand mean, τ i is effect of ith species and ε ij is an error observation. The minimal and maximal values of traits as well as arithmetical means and coefficients of variation-CV (in %)-were calculated. Moreover, the 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 coefficients using FCORRELATION procedure in GenStat 17th edition. Results were also analysed using multivariate methods. The canonical variate analysis was applied in order to present multitrait assessment of similarity of tested species in a lower number of dimensions with the least possible loss of information [45]. This makes it possible to illustrate variation in species in terms of all observed traits in the graphic form. Mahalanobis' distance was suggested as a measure of "polytrait" species similarity [46], whose significance was verified by means of critical value D α called "the least significant distance" [47]. Mahalanobis' distances were calculated for species. The differences among analysed species were verified by the cluster analysis using the nearest neighbour method and Euclidean distances [48]. Difference between values of P/E for monads and P/d for tetrads was tested on the based of t-test. All the analyses were conducted using the GenStat v. 17 statistical software package [49].

Results
The study was conducted on 45 Erica species which represent alle five subgenera and 22 sections of the genus Erica (Table 1). The pollen grains of Erica studied species were usually dispersed as tetrahedral tetrads (39 species), rarely as monads (6 species). Quantitative features of pollen grains are summarized in Table 2 (monads) and in Tables 3-5 (tetrads) and illustrated with scanning electron micrographs (Fig 1 - (Table 2). Mean values and coefficients of variations (CV) for the observed traits indicate high variability among the species for which significant differences were found in terms of all analysed traits ( Table 2).
The mean length of the equatorial diameter (E) was 20.52 (14.00-28.00). The shortest mean equatorial diameter occurred in the pollen of E. nabea (16.53 μm), while the longest was in E. fastigiata (23.93 μm; Table 2).
In all studied taxa, the outline in polar view was mostly circular, more rarely triangular or elliptic, whereas in equatorial view it was mostly elliptic.
Pollen grains usually possess three apertures-colpori. Ectoapertures-colpi were arranged meridionally, regularly, more or less evenly spaced, and were ussualy long; mean length 18.93 (14.00-26.00) μm (Table 2). On average, the length of colpi constituted 75% of the polar axis length. On average, the shortest colpi were found in E. glabella (17.13 μm), while the longest were found in E. fastigiata (21.2 μm). Colpi were fusiform in outline. Their width was variable and usually greatest in the equatorial region. Sculpturing of ectocolpus membranes varies from psilate to granulate. In E. puberuliflora the bridge is present. Endoapertures were usually located in the middle of colpi, less frequently asymmetrically, usually singly, very rarely in Pollen morphology of species from Erica (Ericaceae) genus Pollen morphology of species from Erica (Ericaceae) genus  Pollen morphology of species from Erica (Ericaceae) genus  pairs. They were circular or elliptic in outline with irregular margins. Endoapertures usually distinct; opened or closed. Exine ornamentation can be granulate and at the poles and along the colpori psilate (E. fastigiata), granulate-fossulate (E. plumosa, E. puberuliflora) or psilate with very numerous microgranules with the diameter usually 0.1 μm (E. glabella, E. globiceps) or with numerous microgranules with the diameter 0.1-0.3 μm (E. nabea). On the basis of the pollen size and exine ornamentation all six studied Erica species with monads were distinguished (see: Pollen key).
The performed correlation analysis indicates statistically significant correlation coefficients of 13 out of 21 coefficients (Table 6). In the case of seven pairs of traits we observed positive correlation coefficients: P and E, P and Le, P and Ex, E and Le, E and Ex, Le and Le/P, P/E and Ex/E. This means that a value increase of one trait in a given pair leads to a value increase of Pollen morphology of species from Erica (Ericaceae) genus the second trait. The negative correlation coefficients was observed between: P and Le/P, E and P/E, E and Le/P, Ex and P/E, Ex and Le/P, Le/P and Ex/E.
The greatest variability in all the analysed phenotypic traits expressed jointly with the greatest Mahalanobis distance was observed between the E. fastigiata and E. nabea (5.183) as well as between E. nabea and E. plumosa (4.887) (Fig 9). In turn, the greatest phenotypic similarity (the smallest Mahalanobis distances) was observed for E. fastigiata and E. plumosa (0.808). The first two canonical variables accounted for 88.28% of total multivariate variability between species with monads (Fig 9). This diagram of the first two canonical variables was used to divide the studied species into four groups. The first and second group comprised one taxon E. nabea and E. globiceps. The third group embraced two taxa, E. glabella and E. puberuliflora and to the last group belonging E. plumosa and E. fastigiata. The above data were confirmed by the dendrogram obtained as a result of grouping by using Euclides method (Fig 10). Pollen morphology of species from Erica (Ericaceae) genus
The mean length of the pollen grains equatorial diameter (d) was 25.96  μm. The shortest mean equatorial diameter occurred in the pollen of E. sparsa (18 μm), while the longest was in E. cinerea and E. vestita (36.4 μm) (Tables 3-5). The tetrad diameter (D) is usually greater, than the length of the pollen equatorial diameter (d). D/d ratio was 1.45 (0.77-2.5) and in averange ranged from 1.36 in E. adnata to 1.66 in E. lucida (Tables 3-5).  Pollen morphology of species from Erica (Ericaceae) genus The average length of the polar axis (P) was 19.70 (10-34) μm. Generally speaking, the smallest mean P were found for the pollen of E. erigena (12.33 μm) and the largest for E. vestita 29.2 μm; Tables 3-5).
The apertures form pairs at six points in the tetrad. Three apertures pairs appear at the proximal and distal face of the tetrad. According to Hesse et al. [37] we have adopted that in Erica species occurs colporus, that means compound aperture composed of a colpus (ectoaperture) combined with an endoaperture of variable size and shape. Colpori were arranged regularly, more or less evenly spaced, and were the average length; mean length of colpus (2f) is 19.75 (10.0-42.0) μm (Tables 3-5). On average, the length of colpus (2f) constituted 0.53 of the tetrad diameter (D), which means that colpus is mostly a half shorter than tetrad diameter (Fig  12). This feature (2f/D) wykazuje duży zakres wartości from 0.29 to 0.90. The shortest mean Pollen morphology of species from Erica (Ericaceae) genus colpi were found in E. parkeri (12.13 μm) and the longest in E. vestita (33.87 μm). Colpi were fusiform in outline. Their width was usually slightly greater in the equatorial region. Colpus membrane are psilate to granulate. Endoaperture in colporus were circular or elliptic in outline with irregular margins, with or without costae. Endoaperture are usually located in the middle of colpi, less frequently asymmetrically, usually singly. Opened, single endoapertures occurred in the majority of studied species. In some species (E. cinerea, E. coccinea, E. denticulata, E. hirtiflora, E. regia, E. tenella, E. tetralix) a bridge is also present, understood according to Hesse et al. (2009), as an exine connection between the margins of a colpus in the equatorial region or as an exine connections within tetrads.
In SEM exine ornamentation is usually granulate and granulate-fossulate, rarely fossulate or psilate (see: pollen key). In many species microgranules with diameters from 0.1 to 0.3 μm were found, while in E. caffra there is only a very fine microgranules with diameters 0.1 μm. Their number varies from a few to very numerous. Some Erica studied species create a specific exine ornamentation, that occurs mainly in tetrads (e.g. E. curviflora, E. discolor, E. erigena or E. tetralix) very rarely in monads (E. fastigiata). On the polar areas (in monad) and at the proximal poles of the tetrad and along colpi exine were psilate, and in the interapertural areas granulate, granulate-fossulate or fossulate.
The performed correlation analysis indicates statistically significant correlation coefficients of 23 out of 55 coefficients ( Table 7) Table 7).
The first two canonical variables accounted for 84.48% of total multivariate variability between species (Fig 13). The greatest variability in terms of all the analysed traits expressed jointly with the greatest Mahalanobis distance was recorded for the E. vestita and E. sparsa (18.298). In turn, the greatest phenotypic similarity was observed for E. plumigera and E. zwartbergensis (0.675), E. peziza and E. trichophylla (0.930) as well as for E. hirtiflora and E. melanthera (0.956).
The above data were confirmed by the dendrogram obtained as a result of agglomeration grouping using Euclides method (Fig 14). The studied species were divide to into three groups. The first group comprised one taxon E. vestita, which is most different from all other species. To the second group belonging ten species, from which the most distinctive pollen grains features shows E. baroniana. The third, large group embraced 28 species. E. regia, E. discolor and E. curviflora and E. cinerea and E. monsoniana are similar to each other and differ from the rest of the group. E. erigena occupies also a separate position in the third group. Other species from this group fall into the seven non-significant subgroups.
From 39 analysed Erica species creating tetrads, on the basis of the tetrad diameter and exine ornamentation eight species (E. arborea, E. australis, E. cinerea, E. erigena, E. parviflora, E. passerinae, E. regia, E. tetralix) were separated and other heaths created a small groups, usually containing two or three species, rarely up to a seven species (see: Pollen key).

Discussion
The research results presented here corroborate the opinions of other palynologists (e.g. [6,26,28,35,36,38,39,50]) that the diagnostic features of Erica species pollen grains comprise: the dispersal unit of the pollen grains, pollen and tetrad size, pollen shape (P/E and P/d ratios) and exine ornamentation.
All researchers agree that the dispersal unit is one of the most important features of the pollen grains of the Erica genus. On this basis, Erica species can be divided into two groups which have monads and tetrads. They were analysed [51] 138 Erica species from southern Africa, 125 of which had tetrads and only 13 had monads. Similar proportions were observed in 21 European Erica species of which only E. terminalis and E. spiculifolia (= Bruckenthalia spiculifolia) had pollen in monads [6,20,22]. In addition, in the present study, species with tetrads (39) dominated over those with monads (6). On the other hand, in the minor genera, the majority of the species studied had monads (57) and the remaining ones (30) had tetrads [6].
One of the palynologists [28] reported an interesting result concerning the dispersal unit of the pollen grains of the studied genus. In his opinion, E. glabella created monad and tetrad pollen grains. This may be because of that, the pollen grains of E. glabella are loosely grouped together in tetrads, which are separated easily into monads. Especially, that monad pollen grains of Erica are the most advanced pollen character state and derived from tetrads. Although other [26] claims that no species has been found with monad and tetrad pollen grains in the Erica as well as other genera of the family Ericaceae. In some studies [6], pollen of E. glabella were described as monad, but in [28] this pollen was described as tetrad. The results of the present study did not corroborate the results cited above [28]: the pollen grains of E. glabella were described here as monad. Furthermore, all the remaining Erica species under investigation in this study had either monad or tetrad pollen grains, which would confirm the hypothesis quoted above [26]. However, inside one section, e.g. Callista, both species-forming monads (E. fastigiata) as well as tetrads (E. denticulata) can be observed. Perhaps this interesting problem will be clarified through further palynological investigations of the large and diverse Erica genus.
The mean measurement results for the monads and tetrads were, in general, consistent with those reported by the above-mentioned palynologists, although higher value ranges were generally obtained for the examined features [e.g. tetrad diameter (D)-result from present study 20-64 μm, [35] 25.6-47 μm, [28] and [36] the same results 29.8-48.4 μm; polar axis (P)-result from present study 10-34 μm, [28] 15.5-26.4 μm)]. This may be due to the examination of different species and larger pollen samples.
The present study fully confirmed the hypothesis presented in [36] that the P/E ratio shows distinct differences between the Erica monads and tetrads. The pollen grains are usually elongated in monads and flattened in tetrads. However, the study presented here does not support the view of the above-quoted researchers that exine ornamentation allows a distinction between Erica monads and tetrads because in the current case, in both of these dispersal units, similar exine ornamentation types occurred.
In addition, this study failed to corroborate the importance of the 2f/D ratio (length of ectoaperture to tetrad diameter) on the basis of which two authors [36] distinguished two Pollen morphology of species from Erica (Ericaceae) genus distinctly different palynomorphological Erica species groups, which is somewhat surprising since their range for this feature amounted to 0.43-0.75, smaller than that obtained in the present study (0.29-0.90). It is worth stressing that the mean values of this feature in individual species present a small range (0.39-0.64) which does not allow identification of species groups. A similar situation occurs in the case of the L/P ratio (length of ectoaperture to length of polar axis). The species producing monads analysed in the present study exhibited a narrow range of mean values for this feature (0.67-0.85), therefore, it was not possible to distinguish species groups. That is why, identification among the same species of two distinct groups with an almost identical range of the described feature (0.68-0.84) raises some doubts [36].
The greatest differences in descriptions of the pollen grains of Erica species concerned the exine ornamentation. These resulted from the high variability of exine ornamentation types and subtypes in individual Erica species, as well as from simplified or expanded classifications of features [26,28,35,36]. In the present study, it was possible to distinguish the greatest number of Erica species or species groups on the basis of exine ornamentation traits (see: Pollen key). The present study supports the opinions of other authors [6,26,36] in that exine ornamentation in the Erica species widely differed but was commonly characterised by numerous microgranules. It was exactly the presence or absence, and number of microgranules and exine ornamentation type which proved to be vital in the diagnosis of the Erica species under investigation. The same types of exine ornamentation was found in species with tetrads and monads [26]. The results from the present study were similar; boundaries between the exine ornamentation types in the tetrads and monads were not distinct. In the analysed monads, psilate, granulate, granulate-psilate or granulate-fossulate exine ornamentation occurred, while in the tetrads-apart from those mentioned above-fossulate ornamentation also occurred. In comparison with the present study, in the case of the species examined by two palynologists [28,36], a greater variability of exine ornamentation types was observed. Rugulate, verrucate, striate, rugulate-psilate and gemmate-psilate pollen grains were reported. This can certainly be associated with the greater variability in this feature noted in the Erica species in their research, as well as with a more expanded classification of exine ornamentation types than that adopted in the present study, follows [37][38][39]. Pollen morphology of species from Erica (Ericaceae) genus Very interesting results were obtained following the statistical analysis of the biometrical features of the pollen in the species producing tetrads (Fig 12, Table 1), but these results are not unequivocal. Despite the heterogeneity of the species groups distinguished, they often included, although not always, species belonging to the same subgenera or sections distinguished in accordance with the adopted systematic division of the Erica genus (Fig 12,  Table 1). Species from the Syringodea subgenus and Gigandra (E. coccinea and E. plukenetii) and Evanthe (E. curviflora E. discolor) sections grouped together, whereas species from the Pleurocallis (E. regia, E. vestita) section occupied a separate position. The most numerous subgenus-Euerica, was represented by 14 species belonging mostly to group 3, and less frequently to group 2. Very similar features were found to occur in species of pollen grains which belonged to Ephebus (E. caffra, E. parviflora, E. hirtiflora, E. peziza) and Arsace (E. arborea, E. hispidula) sections, as well as E. trichophylla derived from the Orophanes section. The remaining species differed from one another to a varying degree but the most similar pollen grains were found in E. tetralix and E. umbellate. On the dendrogram, the representatives of the Eurica subgenus which differed from the others the most were E. tenella and E. zwartbergensis, whose pollens were very similar to each other. Finally, four heathers from Madagascar occurred in group 2 (E. baroniana, E. goudotiana) and group 3 (E. cryptoclada, E. parkeri), with the second pair characterised by very similar pollen grains (Fig 12, Table 1). From among the Chlamydanthe subgenus, three species clustered together, two derived from the Trigemma (E. plumigera, E. baccans) section and one from the Elytrostegia (E. diosmifolia) section; the remaining heathers were found in other groups. In the Platystoma subgenus, similar features were found to occur in four (E. melanthera, E. passerinae-section Gamochlamys, E. sparsasection Polycodon and E. adnate-section Eurystoma) out of five of the species examined. Dissimilarity was observed in E. lucida from section Eurystoma. From among the minor genera species, only two formed tetrads, namely E. axillaris and E. totta, which belonged to the same group.
In the case of species producing monads, the results obtained as well as the distribution of taxons acquired on their basis failed to reflect the internal division of the Erica genus generally Pollen morphology of species from Erica (Ericaceae) genus accepted in taxonomy (Fig 14, Table 1). The species producing monads did not create a homogeneous group in the system. Four of them (E. glabella, E. globiceps, E. plumosa, E. puberuliflora) came to the genus Erica from minor genera. They were included in different groups, with the exception of E. glabella and E. puberuliflora (Fig 14). The pollen grains of E. fastigiata were most similar to E. plumosa, although the former species derives from the section Callista, to which E. denticulata producing tetrads also belongs, whereas the latter species belongs to the minor genera. Only E. nabea exhibited discreteness, and belongs to a separate Chlamydanthe subgenus and Adelopetalum section.

Conclusions
The study revealed that the diagnostic features of the pollen grains studied were: pollen dispersal unit, exine ornamentation, P/E ratio, tetrad diameter (D) and length of polar axis (P). On the basis of these traits, 14 Erica species (six creating monads and eight-tetrads) were distinguished which, in the case of pollen features, constitutes a significant number. Other heaths created small groups, usually containing two or three species, but up to seven species. Our study, based on the highest number of Erica species (45) analysed so far, corroborated the view that an examination of palynological features may assist in clarifying classification systems for the large and taxonomically very difficult Erica genus, in particular, at the level of the subgenus and section, but also at species level. The results obtained indicate the need to continue palynological investigations on the Erica genus.