Quantifiable differences between phytolith assemblages detected at species level : analysis of the leaves of nine Poa species ( Poaceae )

The taxonomic value of phytolith assemblages and their degree of variability within different species of the same genus is still an undervalued issue in the botanical range of phytolith studies. However the understanding of grass phytolith variance and its implications to plant systematics is doubtless. In the present study phytoliths of the lateral shoots (leaves) of nine, globally distributed Poa species (Pooideae – Poaceae) are described. Phytoliths were recovered from Poa specimens by the dry ashing technique. Altogether 6223 disarticulated phytoliths were counted (approximately 500–700 phytoliths per species) in 54 plant samples, which cover six shoots of nine species. Not only the relative frequency of each morphotype was calculated, but measurements were conducted to determine the biogenic silica content of Poa lateral shoots. A phytolith reference collection for the nine selected species of a worldwide importance was also compiled. The description of the most significant phytolith morphotypes and their taxonomic relationships are given here. Results suggest that the biogenic silica content of the Poa lateral shoots was determined to be relatively high within all nine species. Phytolith assemblage data was subjected to multivariate statistical analyses (e.g., CA and PCA) in order to find differences and similarities among the nine Poa species. Results show that the two closely related Poa of the P. pratensis species group, namely the P. pratensis and P. angustifolia, only slightly differ from the other Poa species if we consider their rondel-trapeziform short cells (SC) phytolith frequencies.


Introduction
Many plants deposit hydrated SiO 2 in cell walls, cells and characteristic structures in intercellulars (silica bodies, phytoliths) [1,2].Several studies link phytolith morphotypes to plant families and subfamilies (in Poaceae) [3][4][5][6][7][8].In addition, it has been shown that there is a correlation between phytolith morphotypes and the C3 and C4 photosynthetic pathways and this allows the reconstruction of plant communities and climate through time [7,8].It seems though that the phenotypic plasticity of phytoliths induced by environmental changes cannot entirely hide the potential of phytolith morphometries in discriminating plant taxa.Some cases are known in which the phytolith shapes of plant reference material allowed the correct identification to species level [9].For example Mejia-Saules and Bisby [10] found distinguishing silica bodies in the lemmae of Melica species.Form and position of phytoliths, which are not greatly influenced by environmental factors but are genetically controlled, may have considerable systematic potential [11][12][13].
With approximately 500 species, the meadow-grass genus Poa L. is the largest genus in the Poaceae [14], and Poa species are common in the Northern Hemisphere.The genus includes important turf and forage grasses (Arrhenatheretea) [15,16] and cultivars as well as troublesome weeds and invaders [14,17].Although primarily found in the temperate zone, Poa is a very widely distributed grass genus [18].Blinnikov et al. [19] used Poa phytoliths alongside Stipa and Festuca in the identification of vegetation composition in the Columbia basin.Regarding the Poa genus, the Poa pratensis L. species group [P.alpigena (Fries) Lindman, P. pratensis L., P. angustifolia L.] [20] is one of the taxonomically most complicated groups in the world.Poa stiriaca Fritsch et Hayek can also be classified in this group [15].The features of the leaf epidermis, and the presence and distribution of silicified unicellular hairs and prickles were shown to be useful characteristics to separate and identify the species of the P. pratensis species group [21][22][23][24].Nevertheless P. pratensis was often used as a "model" organism in many studies [25].
The aim of this study is to describe the phytolith assemblage in lateral shoots of nine Poa species, looking for potential species level phytolith characters that could be useful in palaeocecology and, in parallel, to start collecting reference material to make comparisons with other grass species and genera.

Sample collections
The following Poa species from the P. pratensis group were collected and analyzed: P. alpigena (Fries) Lindman, P. angustifolia L., P. pratensis L., and P. stiriaca Fritsch et Hayek.Other selected species consisted of two annual taxons: P. annua L., P. compressa L., P. trivialis L. and three perennial taxons: P. arctica R. Br. and P. hybrida Gaudich.The individuals of the species that were selected to be studied are characterized by wide geographical distribution.Most of the plant material originated from caryopses of the nine species obtained from the National Germplasm System of the United States Department of Agriculture [26] Poa collection.They were established in a climate room then planted into experimental plots under the same conditions.The studied germplasm were selected from different countries, with large geographical variation to represent different genetic backgrounds (Tab.1).Proper identification of the grown specimen were undertaken in order to analyse relevant phytolith morphologies.A smaller part of plant material was collected at Hungarian localities (Tab.1), except the specimens of P. arctica, P. alpigena and P. hybrida, because of the areas of these species do not reach the Carpathian basin (Tab.1).

Laboratory procedures and data analysis
Six lateral shoots (leaf blades and leaf sheaths) of each species were compounded and treated as one single sample.Phytolith extraction and biogenic silica content (bSi) measurements were accomplished through the dry-ashing technique based on the methodological guidelines published by Albert and Weiner [27] and Mercader et al. [28,29].The ashes were mixed thoroughly, then mounted on light microscope slides in immersion oil and observed with a Zeiss Axioskop 2+ microscope at a magnification of 1000×.500-600 pieces of identifiable plant opal particles -phytoliths -per species were counted in adjacent but not in overlapping lines across the cover slip (with 22 mm length).Their morphological classification was accomplished based on Twiss et al. [3], whilst the denomination of individual morphotypes was accomplished according to the International Code for Phytolith Nomenclature 1.0 (ICPN 1.0 [30]).The classification and nomenclature work was complemented with the system of other authors (e.g., [16,28,[31][32][33][34]).Meanwhile we documented the phytolith morphotypes by taking microphotographs.The frequency of individual morphotypes was calculated and the results visualized with the help of C2 ecological data visualization software [35].The ubiquity of morphotypes was noted.Correlations between classified phytolith number and biogenic silica content were determined using the Pearson correlation coefficient.Hierarchical cluster analysis based on phytolith assemblages was used to estimate taxonomic value of phytolith morphotypes in the species.The dissimilarity coefficient was Euclidean distance and sorting strategy was single link [36].To establish the degree of segregation of species based on phytolith morphotypes and to analyze the importance of morphotypes in this segregation, principal component analysis (PCA) was performed using SPSS 15.0 for Windows (SPSS, Chicago, IL, USA).
The adaxial and abaxial epidermis of three, dried leaves from three different lateral shoot per species were studied using AMRAY 1830I type scanning electron microscope (SEM).To illustrate the locality of the phytoliths in the epidermis tissue black scatter SEM images were made on P. pratensis and P. angustifolia leaves surfaces.Tab. 1 Inventory of the examined Poa species and their biogenic silica content expressed in the plant's dry weight.

Biogenic silica content
The ash of the Poa shoots contained fully silicified silica bodies, cells with silicified walls, corroded silicified cells and debris of silicified cell walls and parts of the epidermis tissue.The average silica content was 16.8% of the plant's dry weight (range: 7.1% of P. arctica and 29.2% of P. alpigena) in the Poa shoots (Tab.1).There may be some correlation between the bSi content and width of the leaves of the studied species because the higher bSi content was coupled with narrower leaves, but the negative correlation is not significant (r = -0.45;P = 0.22).

Descriptive results of lateral shoots phytoliths in Poa species
Approximately 300-1000 microphotographs were taken of every species with a total of 2703 (66-855 per species) showing classified and a total of 3520 (82-1243 per species) showing unclassified phytoliths.Altogether 6223 phytoliths were counted.Some (142) tissue microphotographs were taken because the cells strongly adhered to each other in the ash (cf.silica sheet elements).Altogether 31 morphotypes are reviewed in this study.This includes those two morphotypes, which are considered to be new [24], because they may bear taxonomical properties.

Classified phytoliths
An average of 76.8% of the nine Poa species phytoliths were identifiable as an established morphotypes (Tab.2).In P. angustifolia, P. stiriaca, and P. hybrida the percentage of classified phytoliths exceeded 60% and in P. compressa it was 50%.This value was below 50% in P. trivialis, under 30% in P. alpigena and P. annua, and little more than 30% in P. arctica.Poa angustifolia had more classified phytoliths and biogenic silica content than the close relative P. pratensis with wider leaves.We did not find correlation between classified phytolith number and biogenic silica content (r = 0.128; P = 0.742).
Tab. 2 and Fig. 1-Fig.5 give a summary of the relative frequencies and observed features of phytolith assemblages of Poa shoots.The phytolith assemblage of the nine examined Poa species leaves were characterized by elongate psilate proportion above 45%, elongate sinuate proportion above 10% and proportion of rondel trapeziform above 20%.Globular, tabular and lanceolate phytoliths were below 5% (Fig. 1).

Phytoliths with high frequency values (above 20%)
The anatomical origin of most of the phytolith morphotypes is the epidermis, as is usual in the Poaceae family.The elongate long cells (LC) and short cells (SC) were common in these species.Elongate morphotypes could be observed with several textures and ornamentation types.The number of ornamentation types varied between two and ten.Poa hybrida and P. alpigena had two elongate ornamentation types, whilst P. pratensis showed ten different ornamentation types (Tab.2).The most common ornamentation types are psilate and sinuate textures, and these can be found in every species.Elongate echinate morphotypes are also common and can be found in leaves of six species.Some of the elongate psilate morphotypes derive from the vascular system (Fig. 4).The elongate depressed psilate morphotype could be found in five species at low frequency, seeming to be a peculiarity of Poa species [24].

Phytoliths with low frequency values (below 20%)
Globular psilate forms with inclusion occurred in leaves of P. alpigena, P. arctica and of P. compressa at high frequency.These phytoliths are presumably mesophyll cells with more or less regular, square inclusions in them.These small square inclusions can also be found in the ash in free state.
Lanceolate trichomes were seen in every species with the exception of P. alpigena.Lanceolate trichomes and bulliforms were in low abundance in this species, the frequencies of trichomes were 0-4.8% and bulliforms were 0-2.8%.Papillae (silicified prickles with rounded ends) were seen only in the leaves of P. pratensis.The trigonal pyramid morphotype was described in P. pratensis individuals [24] and it was found in P. angustifolia and P. arctica leaves at low frequency.Papillae with pitted edges could be found only in P. compressa with a frequency of 0.2%.The lobate morphotypes are not so typical in the studied Poa species (polylobate: 0.1%, bilobate: 0.03%).

Differencies of phytolith assemblages among the nine Poa species
Scanning electron microscope images (Fig. 6-Fig.9) illustrate the phytolith assemblage of the species summarized in Fig. 1.At every species, the adaxial surfaces contain trichomes (prickles) without distinguishable costal zones and silicified short cells.The abaxial surfaces contain distinguishable costal and intercostal zones and costal zones may be abundant in silicified short cells.Poa pratensis (Fig. 6a,b) and P. angustifolia (Fig. 6c,d) back scatter SEM images show the locality of the phytoliths in both of the leaf surfaces by the lighter silica bodies.The adaxial epidermis produce more silicified long cells (and trichomes) than the abaxial surfaces.Stomata are not silicified.The leaf epidermis of P. pratensis can be characterized by abundant rondel-trapeziform SC and lanceolate unicellular trichome (or short prickle) phytoliths.The adaxial surfaces produced more trichomes and elongate LC morphotypes than the abaxial surfaces (Fig. 7a,b).Epidermis surfaces of P. angustifolia leaves are similar to those of P. pratensis but costal zones are wider with more cell rows than the costal zone of adaxial epidermis of P. pratensis leaves (Fig. 7c,d).Silification of short cells was not so typical in P. alpigena epidermis (embedded picture in Fig. 7e,f), but globular inclusion and small square phytoliths were observed in bigger amounts (Fig. 1).Poa stiriaca was characterized by abundant elongate LC, specially high elongate castelate LC, and the least rondel-trapeziform SC among the studied Poa species (Fig. 8a,b).Poa annua is characterized by rectangulare phytoliths, but prickles were not found to be typical for this species (Fig. 8c,d).Poa arctica leaves produced more globular morphotypes than the other studied species.Rondel-trapeziform short cell could have been observed  with different sizes in surfaces of P. arctica leaves (Fig. 8e,f).There are some silicified SC in the costal zone of abaxial epidermis of P. annua and P. arctica but it does not reach the number accordance with the abundant trichomes of silicified SC of P. angustifolia or P. pratensis.Poa compressa leaves characterized by less rondel-trapeziform SC and the most number globular inclusion phytoliths (Fig. 9a,b).Lanceolate was the dominant morphology in P. hybrida, which is in accordance with the abundant trichomes (Fig. 9c,d).There were lot of elongate LC with different ornaments in P. trivialis.As Fig. 9e,f shows, scutiforms -originated in unicellular hairs -were also common.Poa hybrida abaxial epidermis produce more rondel-trapeziform SC than the P. compressa and P. trivialis does.Cluster analysis revealed four main groups: group 1, consisting of P. pratensis and P. angustifolia; group 2, consisting of the other species belonging to the P. pratensis species group: P. alpigena, P. stiriaca, and P. trivialis (the latter species is not a member of the group); group 3, consisting of three species: P. annua, P. arctica and P. hybrida; and group 4, consisting of P. compressa (Fig. 10).
The two closely relative species in group 1, P. pratensis and P. angustifolia, were characterized by more rondeltrapeziforms SC phytoliths and fewer elongate psilate LC than in the other Poa species (Fig. 1).Although group 3 (P.annua, P. arctica and P. hybrida) produced abundant rondel-trapeziform phytoliths, the rondel-trapeziform content is about half that of group 1.At the same time group 3 had about twice as many elongate LC morphotypes than group 1 (Fig. 1).The phytolith assemblage of P. alpigena was to a certain degree separated from the others in group 2. It is characterized by globular inclusion and square morphotypes.The phytolith assemblage of the two other species in group 2, P. stiriaca and P. trivialis, were similar to each other.Group 4 consisted of only P. compressa, which was characterized by globular inclusion phytoliths (Fig. 1).
The first two axes of the PCA (Fig. 11) amounted to 97.6% of the total variance (67.0% for axis 1, 30.5% for axis 2).On Fig. 11, where axes 1 and 2 are represented, two groups can be noted: group I: P. pratensis and P. angustifolia; species belonging to the P. pratensis species group); group II: other species belonging to the P. pratensis species group: P. alpigena, P. stiriaca, and the species that are not group members are P. trivialis, P. annua, P. arctica and P. hybrida, P. compressa.The morphotypes that mainly contribute to axis 1 are different types of elongate psilate phytoliths; rondeltrapeziform phytoliths contribute to axis 2. PCA of phytolith morphotypes confirms this observation (Fig. 12).Axes 1 and 2 are represented in Fig. 12, where axis 1 represents phytolith amount.The proportions of elongate sinuate and elongate psilate morphotypes are high in every species.Standard deviation of abundant rondel-trapeziform phytoliths are represented by axis 2. This morphotype is special because its proportion is high in every species but the standard deviation is higher than that of abundant elongate phytolith morphotypes.

Discussion
Compared to previously conducted studies on the biogenic silica content of Poa species, we can state that the biogenic silica content (16.8%) of the studied species is relatively high (cf.[28,37]), therefore it is anticipated that the Poa genus is present in fossil phytolith assemblages.Poa pratensis shoots (leaf blades and sheaths) accumulate silica mainly in their epidermis cells, both in short and long cells, as this feature is typical in most Poales [13,38].There may be some connection between the bSi content and width of leaves of studied species because Poa specimens with wider leaves had lower bSi contents.The phytolith assemblages of the nine Poa species are compounded by several morphotypes described previously for other species of the family [16], complemented by the two morphotypes found in Poa leaves by Lisztes-Szabó et al. [24].The Poa genus (represented by the nine studied species) is characterized by elongate psilate, rondel-trapeziform phytoliths, and lanceolate or scutiform morphotypes (long, unicellular hairs and short prickles).
There is a diverse range of silica body morphologies in Poaceae, including dumbbell-shaped, cross-shaped intermediates between these two types, horizontally elongated shapes with psilate or sinuous outlines, saddle-shaped, conicalshaped and numerous others [13,38,39].It is interesting to note that the cross-shaped and saddle-shaped forms were absent in the examined Poa shoots, similarly to Brown's results [4], and at the same time lobated phytoliths were not found, unlike in Poa alpina epidermis where this type is dominant [37].Only a few prickles (unicellular trichomes), and especially the peak of them could be observed, because long, multicellular trichomes are not typical in this species [38].However, it is also true that the silica content is higher in the apex of the prickles than in their base, therefore the recovery and observation of apex fragments has a higher probability.
Similarly to the study of Morris et al. [40], only scarce amounts of bulliforms were found.Nonetheless, strongly silicified bullifoms are frequently detected in the soil-phytolith assemblages of the geographical area of these species.It is likely that, the silification of the bulliforms are typical for the C4 species rather, than the C3 Poa species.
The results match and corroborate with the descriptions published by other authors, who noted the predominance of the same phytoliths morphologies: P. secunda ([40] -Great Basin, USA; [16] -Pacific Northwest, USA).The three-part division [3] seems to be relevant for the Poa genus because frequencies of saddle (typical of Panicoideae) and bilobate (typical of Chloridoideae) were low (only one lobate was found in this study), but circular/oval rectangular (in the present study: rondel-trapeziform) frequency was high (typical of Pooideae).
Having summarized our results regarding morphotype frequency, we supported the conclusion of Brown [4] and Marx et al. [41].They argued that interspecific variations of phytolith morphotypes can be observed.Not all shapes common to a given species were found in each specimen.Genetic variation among plants and the geographical location where the plants grew may cause intraspecific differences in phytolith assemblages [42,43] and these may fix at species level.The two closely relative Poa species of the P. pratensis species group, P. pratensis and P. angustifolia, slightly differ in the abundance of rondel-trapeziform SC phytoliths from those of the other Poa species.This result confirmed the previous taxonomic statements, because the phytolith that originate from SC are genetically fixed and have taxonomical relevance [44,45].Blinnikov et al. [46] also identified several rondels and plates of blue meadow grass in grasslands of controlled composition on experimental plots (at Cedar Creek, Minnesota).Rondels (or truncated cones) have been assigned generally to the subfamily Pooideae.However, there are some pooid species (Melica brasiliana) that do not produce this morphotype, whereas species from other subfamilies produce them in great quantities (Stipa spp.and Piptochaetium spp.).Barboni and Bremond [47] studied 184 East-African grass species and found that there are morphological variations (of size and number of lobes) within the rondels, which could additionally be considered to improve environmental and taxonomical interpretation of phytolith assemblages.
Similar establishment was reported for palms, because it was not possible to identify unambiguously to taxon individual palm phytoliths.Interspecific differences in phytolith morphology significantly outweighed intraspecific variation, revealing the potential value of further research [48].Thorn [49] compared phytolith assemblage of P. litoralis with those of an unidentified Poa species and reported about the variation within the Poa at the generic level.
The cluster analysis and PCA reflect quite clearly the systematics of the group of species under study.Firstly, it is possible to distinguish between the two groups analyzed: P. pratensis agg.and other Poa species.The differences lie in the phytolith morphotypes.As Shaheen et al. [50] found in the genus Brachiaria, and subsequent upon our Poa study the quantitative analysis of phytoliths may provide a new taxonomic character to distinguish different species of a grass genus.

Conclusions
This description of leaf phytolith assemblages of P. pratensis, P. angustifolia, P. alpigena, P. stiriaca, P. trivialis, P. annua, P. arctica, P. hybrida and P. compressa represents the first contribution for nine different species in the same genus.Since Poa species are characterized by diverse geographical distributions it is proposed that their phytoliths are of significant value in ecological studies.Finally, some characters of phytolith assemblages are described to allow the collection of reference material to start in order to compare other grass species and genera in the grasslands of the Northern Hemisphere in the future.Although our results do match the general patterns of phytolith assemblages within a grass subfamily as described by other authors, it is necessary to consider the exceptions found here, especially when linking the relative frequencies of morphologies to plant systematics.

Fig. 7
Fig. 7 SEM images of lateral shoots of Poa species.a Adaxial leaf epidermis of P. pratensis.b Abaxial leaf epidermis of P. pratensis.c Adaxial leaf epidermis of P. angustifolia.d Abaxial leaf epidermis of P. angustifolia.e Adaxial leaf epidermis of P. alpigena.f Abaxial leaf epidermis of P. alpigena.elc -epidermal long cell; esc -epidermal short cell; p -prickle; uh -unicellular hair.

Fig. 8
Fig. 8 SEM images of lateral shoots of Poa species.a Adaxial leaf epidermis of P. stiriaca.b Abaxial leaf epidermis of P. stiriaca.c Adaxial leaf epidermis of P. annua.d Abaxial leaf epidermis of P. annua.e Adaxial leaf epidermis of P. arctica.f Abaxial leaf epidermis of P. arctica.elc -epidermal long cell; esc -epidermal short cell; p -prickle; st -stomatal complex; uh -unicellular hair.

Fig. 9
Fig. 9 SEM images of lateral shoots of Poa species.a Adaxial leaf epidermis of P. compressa.b Abaxial leaf epidermis of P. compressa.c Adaxial leaf epidermis of P. hybrida.d Abaxial leaf epidermis of P. hybrida.e Adaxial leaf epidermis of P. trivialis.f Abaxial leaf epidermis of P. trivialis.elc -epidermal long cell; esc -epidermal short cell; p -prickle; st -stomatal complex.

Fig. 11
Fig. 11 Principal component analysis axes score plot of species based on their phytolith assemblages.