Grass pollen surface ornamentation is diverse across the phylogeny: Evidence from northern South America and the global literature

The grasses are one of the most diverse plant families on Earth. However, their classi ﬁ cation and evolutionary history are obscured by their pollen stenopalynous (similar) morphology. A combination of high ‐ resolution imaging of pollen surface ornamentation and computational analysis has previously been proposed as a promising tool to classify grass taxonomic boundaries. In this study, we test this hypothesis by studying Poaceae pollen across the phylogeny from plants collected in northern South America and also from published literature across the globe. We assessed if morphotypes that we establish using descriptive terminology are supported by computational analysis, if they vary along six (a)biotic variables and vary across the phylogeny. Based on this analysis, we constructed a reference framework for pollen surface ornamentation morphotypes. Our results showed that there is a wide variation of grass pollen surface ornamentation. We identi ﬁ ed nine new and con ﬁ rmed six known morphotypes, establishing a data set for 223 species (243 individual plant specimens) that represent 11 subfamilies. Computational analysis showed that our morphotypes are well ‐ supported by two quantitative features of pollen sculptural elements (size and density). The speci ﬁ c data set and mapping of the phylogeny con ﬁ rmed that pollen morphological sculpture is unrelated to (a)biotic variables and is diverse across the phylogeny.

Pollen morphology has been regarded as an ideal method when studying the microbotanical fossil record (e.g., Ferguson et al., 1977;Page, 1978;Beug, 2004;Mander et al., 2013), and can provide important information on taxonomic boundaries and classification.Nevertheless, it is difficult to classify fossil grass pollen in terms of extant genera or even subfamilies, as there is no complete Poaceae pollen reference data set of extant species.Moreover, Poaceae pollen is stenopalynous (principally monoporate and psilate) and, therefore, very similar within the family (Page, 1978;Salgado-Labouriau & Rinaldi, 1990;Bush, 2002;Beug, 2004;Halbritter et al., 2018), although there have been reports of grass pollen with two to three pores rather than the usual single pore (Mercuri et al., 2022).Recently Wei et al. (2023) showed that the pollen size of undomesticated grasses cannot easily be used to reconstruct past vegetation or climate parameters.In addition, glycerin jelly, a commonly used mounting medium in palynological processing procedures, has been found to affect pollen size during pollen storage (Wei et al., 2023).This all means that it is challenging to reconstruct the history of grasses using simple characters that can be measured with transmitted light microscopy.
In spite of the stenopalynous of grass pollen morphology, since the 1970s, a high degree of variability in grass pollen surface ornamentation has been observed when using scanning electron microscopy (SEM) (e.g., Andersen & Bertelsen, 1972;Page, 1978;Köhler & Lange, 1979).Scientists have made several attempts to classify grass pollen based on this variability.Page (1978) tried to recognize exine patterns based on granules being fused or separated but found that they could not build a classification based on such features.Köhler & Lange (1979) first proposed that the pollen surface ornamentation can be classified into different morphotypes, such as Hordeum-type, Triticum-type, Avena-type, and Setariatype.Subsequently, a new Pariana-type and Stipa-type were identified (Mander & Punyasena, 2015).Nevertheless, the commonness or rarity of each morphotype remained unclear due to the low number of documented specimens with a clear surface pattern.Analyzing pollen morphology under electron microscopy and superresolution microscopy with computational image analysis is now a developing tool that can resolve these problems (Mander et al., 2013;Romero et al., 2020).
Special ornamentation structures in grass pollen might be useful as an additional indicator to understand the fossil record.For instance, the "rugulate" shape and the "orbicular" ornamentation pattern are two morphological forms that have been previously documented (e.g., Vinckier & Smets, 2001;Dórea et al., 2017;Guimarães et al., 2018;Ruggiero & Bedini, 2018).The "rugulate" shape consists of elongated irregular ridges, whereas "orbicules" consist of distinctive granules (Punt et al., 2007).Souza et al. (2021) confirmed that the rugulate ornamentation of Dendrocalamus asper (Schult.& Schult.f.) Backer ex K. Heyne is only shown during pollen maturation, which is related to flowering inflorescence development.Furthermore, Lovisolo & Galati (2012) proposed that the occurrence of orbicules might be different in different subfamilies, and they defined orbicules in the subfamily Pooideae as spherical corpuscles with a spiculated surface, Chloridoideae as plate-like corpuscles, and Panicoideae as subspherical corpuscles, respectively.The orbicular ornamentation pattern is also relevant in the field of allergy studies.It was hypothesized that orbicules make grass pollen an important vector as an allergen (Vinckier & Smets, 2001).Therefore, exine structure could be a factor when attempting to understand allergies in relation to pollen.
In this study, we ask the following questions: (i) Can pollen morphotypes help us to differentiate and classify Poaceae at different taxonomic levels?(ii) Can such classification be supported by quantitative morphometric data of surface ornamentation? and (iii) Does the pollen surface ornamentation vary in concert with key explanatory parameters, or systematically across the grass phylogeny?To achieve this, we took SEM images on the pollen surface ornamentation for 803 pollen grains, which were collected from northern South America from 68 plant specimens belonging to 55 species from nine subfamilies across the Poaceae phylogeny.We also reviewed the literature and identified additional morphotypes from published SEM images of grass pollen surface ornamentation from 170 species (175 specimens).Based on our SEM imagery data, we constructed a database of 15 surface ornamentation morphotypes in total (including five types from published literature).The database helped us assess how these morphotypes are distributed among the grass taxa.We also extracted the size and density of sculptural elements on the surface of grass pollen grains.We used image processing to test the degree to which the morphotypes in our database are supported by quantitative measurements of morphology.We also correlated grass pollen surface ornamentation with explanatory variables, and we mapped morphotype occurrence onto the grass phylogeny in order to explore how pollen sculpture ornamentation varies across the Poaceae phylogeny.

Materials and methods for our samples 2.1.1 Sample collection and processing
The pollen grains were sampled from 55 species of 68 plant specimens from the Naturalis Biodiversity Center of the Netherlands (NHN-L).These specimens were all deposited in the collection between 1808 and 2012 and were all collected from northern South America (Fig. 1).Two to three specimens from eight species belonging to six subfamilies were selected here for testing the variability within species, including Aristida recurvata Kunth, Calamagrostis effusa (Kunth) Steud., Chusquea tessellata Munro, Echinochloa colona (L.) Link, Paspalum repens P.J. Bergius, Poa annua L., Sporobolus indicus (L.) R.Br., and Streptochaeta spicata Schrad.ex Nees.The specimen information (country and year, collector information, location, elevation, photosynthetic pathway, climate data, vegetation type, and soil type) can be found in Appendix S1.
All pollen samples were available through previous research and methods are reported in Wei et al., 2023).In the present study, we transferred the pollen from glycerin jelly to ethanol with a gradual concentration series (70%, 96%, and 100%), to remove the glycerin jelly and dehydrate the grains ahead of SEM imaging.

SEM image acquisition
Dehydrated pollen grains were transferred to SEM stubs and coated with Platinum gold, then scanned with SEM (TESCAN MAIA3, Czech) under 20 kV at the Nanjing Institute of Geology and Paleontology, Chinese Academy of Sciences (NIGPAS).At least, nine pollen grains were examined and imaged for each plant specimen.Each pollen grain was imaged under three amplification levels: ×15 000, ×75 000, and ×150 000.All images were saved in high resolution for morphological analysis.

Palynological morphotype description of Poaceae
The pollen surface ornamentation was described based on the SEM images and each specimen was grouped into one specific morphotype.The descriptions of new morphotypes follow Punt et al. (2007).Six established known morphotypes refer to previous work (Köhler & Lange, 1979;Mander & Punyasena, 2015).

Quantifying pollen surface ornamentation with size and density of sculptural elements
Quantitative analysis was undertaken on 3.3 μm 2 (1855 × 1855 pixel) windows that were manually cropped from the ×150 000 images (Figs.2A1, 2B1).Size and density were extracted from the pollen surface ornamentation of each SEM image using the method adapted from the approach developed by Mander et al. (2013).First, contrast-limited adaptive bell-shaped histogram equalization was used for adjusting the image contrast (Figs.2A2, 2B2).Subsequently, five subimages with 1000 × 1000 pixels were randomly chosen and cropped from each 1855 × 1855 image.For each subimage, Sobel approximation was applied to detect edges with maximum gradient, then morphological closing (a dilation process followed by an erosion process) was performed to connect scatter points (Figs.2A3, 2B3).We then identified the first 10 largest connected components.For each connected component, we performed principal component analysis (PCA) and calculated the value of PC1 projection as its size.We took the average size for the first 10 largest connected components as the size of each subimage (Figs.2A4, 2B4).Finally, we took the average of the size features for the five subimages as the first feature [size] of one 1855 × 1855 image.
Afterwards, each subimage was quantized into four colors by minimum variance quantization and the brightest color was used as the foreground (i.e., binarize the image, Figs.2A5, 2B5).We then performed morphological closing on the binary image to connect the scatter points and counted the connected components with the areas larger than a threshold as the density of the subimage.We then took the average of all density features for the five subimages; the second feature [density] was obtained (Figs.2A6, 2B6).
The value of size and density feature of each grain of northern South American samples are provided in Appendix S2.
The vegetation type, soil type, elevation, mean annual temperature (MAT), and total annual precipitation (TAP) data for each of our northern South America samples were taken from Wei et al. (2023).
2.1.5.2Data visual exploration.The size and density values from the image processing step were transferred to log size and log density before being used in data analyses, which aimed to make the data closer to a normal distribution.Box plots, scatter plots, and linear regression were used to explore and visualize the data.
2.1.5.3 Software, SEM images, and data availability.All image processing and feature extraction were carried out in MATLAB (R2017a).All data analysis and visualization were carried out using R v.4.1.1 (R Core Team, 2021).All original SEM images of grass pollen ornamentation for extracting quantitative features, the MATLAB code for quantitative analysis, and the R code for data analysis are available for download from figshare: https://doi.org/10.6084/m9.figshare.23302022.v3

Results
3.1 Poaceae pollen morphotypes 3.1.1Our samples Based on our study of surface pollen ornamentation of 68 grass specimens, we assigned these specimens to 10 morphotypes (Appendix S1).Four of these 10 types are new morphotypes, including Arthrostylidium-type, Muhlenbergia-type, Poa-type, and Streptochaeta-type (Fig. 3A; Table 1).All six known morphotypes that were described in the earlier literature (Köhler & Lange, 1979;Mander & Punyasena, 2015) were found in our samples (Fig. 3B; Table 1).The descriptions of all morphotypes are provided in Appendix S5.Note that Arthrostylidium-type, Muhlenbergia-type, Poa-type, and Streptochaeta-type are new morphotypes, and their scanning electron microscope (SEM) images can be found in Fig. 3A.Avena-type, Hordeum-type, Pariana-type, Setaria-type, Stipa-type, and Triticum-type are known morphotypes, and their SEM images can be found in Fig. 3B.
SEM images of grass pollen surface ornamentation.We identified 15 morphotypes in total (Appendix S7), and among them, the Avena-type, Stipa-type, Hordeum-type, and Setariatype are the most common.Five of these 15 types were identified as new morphotypes, including Anomochloa-type, Diandrolyra-type, Olyra juruana-type, Olyra bahiensis-type, and Sucrea-type (Table 2).The descriptions of these five new morphotypes are organized in Appendix S5.
3.2 Ornamentation variation within species 3.2.1 Our samples Eight species were chosen for testing the variation of pollen ornamentation among different specimens within species.Seven species (87.5%), present variability of the pollen surface ornamentation among different specimens (Fig. 4).For example, although three specimens of Chusquea tessellata were all collected from a tropical moist forest vegetation type in Colombia, their morphotypes are quite different, including Avena-type, Triticumtype, and Hordeum-type (Figs. 4B1-4B3).This is the first time that within-species variation in the pollen exine sculpture is reported for the Poaceae.

Literature study
For five species, multiple plant specimens collected from different locations were found in different publications (Appendix S7).We found that pollen surface ornamentation varied in two of these five species (Table 3).The Stipa-type and the Hordeum-type were found within Setaria verticillata (L.) P. Beauv.(Morgado et al., 2015;Ullah et al., 2021).Within-species variation was also found in domesticated crops such as Zea mays, with the Hordeum-type and Stipa-type both observed in this species (Watson & Bell, 1975;Morgado et al., 2015).Moreover, the pollen ornamentation of two of our South American species, Streptochaeta spicata and Olyra micrantha Kunth, was earlier reported (Page, 1978;Mander & Punyasena, 2015;Dórea et al., 2017), and additional morphotypes for each species were identified.Specifically, the Hordeum-type and Setaria-type were found in S. spicata (Dórea et al., 2017) and O. micrantha (Page, 1978).

Overview of pollen surface ornamentation types in the Poaceae
By combining our samples and published SEM data, we constructed an updated database for grass pollen surface ornamentation, which includes 243 specimens from 223 species (Appendix S7), which were distributed in a wide variety of climates and environments.These samples represent 11 subfamilies across the grass phylogeny.A total of 15 morphotypes were identified, with nine of these characterized for the first time and added as new types, including the Arthrostylidium-type, Anomochloa-type, Diandrolyra-type, Muhlenbergia-type, Olyra juruana-type, Olyra bahiensis-type, Poa-type, Streptochaeta-type, and Sucreatype.Nine of the 10 morphotypes that are present in our South American samples were found in the published literature.In general, the Avena-type, Setaria-type, Hordeum-type, Stipa-type, and Triticum-type are most common (Fig. 6).

Correlation between human-defined pollen morphotypes and quantitative analysis
Box plots were used to explore the relationship between the two quantified features (size and density) of pollen surface ornamentation and the human-defined morphotypes.Our results showed that each morphotype represents a unique morphological space of size or density, despite some variability that leads to an overlap between morphotypes.For Stipa-type, Hordeum-type, Triticum-type, Avena-type, and Poa-type, their log size and density both show a substantial increasing trend, and the log density value of Poa-type reached a peak.After the peak at the Poa-type, the value of log density of Setaria-type, Arthrostylidium-type, Muhlenbergia-type, and Pariana-type started to decrease.The Streptochaeta-type is the only one with both relatively large size and density (Figs.7A, 7B).We conclude that humandefined morphotypes are well supported by computational analysis of the size and density features of sculptural elements.

Correlation between pollen ornamentation features and explanatory variables
Box plots, scatter plots, and regression were used to explore the correlation between pollen ornamentation features (size Table 4 Morphotypes with rugulate structure from existing publications.

Species
Morphotypes References and density) and six explanatory variables (Figs. 8, 9; Appendix S6).The distribution of log size and log density across each box plot of these six variables showed a substantial overlap with similar median values.Our results suggest that pollen ornamentation does not show a clear relationship with vegetation type, soil type, photosynthetic pathway, elevation, MAT, or TAP.

Overview of our results
The main focus of our work was to explore the variation in pollen surface ornamentation and to determine if there is a correlation with their taxonomic position in the Poaceae.Scientists previously grouped grass pollen surface types into  six morphotypes (e.g., Köhler & Lange, 1979;Mander & Punyasena, 2015).However, this number may not represent the entire pollen morphotype variation within the Poaceae.
In addition, we question if morphotypes have taxonomic and paleoecological relevance beyond their original use with domesticated grasses.In our study, we aimed to address this problem through extensive sampling across northern South America and the use of data collected from other localities reported in the literature.We identified nine new and six known pollen surface morphotypes in total.These types are well supported by quantitative analysis of the size and density of sculptural elements of the pollen surface.We revealed that grass pollen surface ornamentation is diverse across the phylogeny.

Is qualitative analysis useful and what does it tell us?
Until now, researchers found that Poaceae pollen surface ornamentation has a high degree of variability that can be classified into different morphological groups or morphotypes (Köhler & Lange, 1979;Mander & Punyasena, 2015).However, there is no assessment of the degree to which each morphotype can be supported by quantitative analyses of the surface ornamentation.In our study, we aimed to assess whether quantitative analysis supports these and other (new) pollen ornamentation morphotypes, when applied to extensive sampling.Ten morphotypes were identified by the human eye from 68 samples collected from different environments in northern South America.Furthermore, after extracting two features of the ornamentation (size and density) based on the SEM images, we found that each morphotype represents a unique morphological space of size or density, despite some variability which leads to overlap between morphotypes (Fig. 7).A trend that is consistent with the results observed by the human eye level, that is, different morphotypes are partially related, and partially different from each other.Therefore, we conclude that human-defined morphotypes are well supported by computational analysis of the size and density features of sculptural elements, and change following a sliding scale trend.We also evaluated the relevance of the quantitative characters of these ten morphotypes with several key explanatory variables, including vegetation type, soil composition, photosynthetic pathway, elevation, temperature, and precipitation.However, there is no clear relationship between them.
When linking pollen surface ornamentation to the dated phylogenetic tree, we notice that the amount of pollen morphotypes identified in our study correlates with the number of species in a subfamily (Fig. 10).The ancient subfamilies (Anomochlooideae, Pharodieae, Puelidideae) that are rooted in the early Cretaceous (101 Ma) (Huang et al., 2022), are low in species richness and have also limited morphological types.In contrast, morphotypes from PACMAD and BOP clades are much more diverse, which split from each other at 81.43 Ma and became hugely species-rich in the Miocene (<23 Ma; Stromberg, 2011;Gallaher et al., 2022;Huang et al., 2022).From this, we conclude that grass pollen morphology is diverse across the phylogeny.

Four common morphotypes throughout the grass phylogeny
We noted that there are four common morphotypes, which already were identified in the 1970s (Köhler & Lange, 1979), including Avena-type, Hordeum-type, Setaria-type, and Triticum-type.These morphotypes are found in the ancient subfamilies (Anomochlooideae, Pharodieae, Puelidideae) and the PACMAD and Bambusoideae-Oryzoideae-Pooideae (BOP) clades.These subfamilies and their clades are distributed across elevations and in different climates, soils, and environmental systems (Gibson, 2009;Blair et al., 2014;Linder et al., 2018), which also explains why the morphotypes, which are delimited by small variations in the size and density of the sculptural elements, are not obviously correlated with the tested explanatory variables (vegetation type, soil composition, photosynthetic pathway, elevation, and climate data) (Figs.8-10).

Unique morphotypes point at taxonomic position
According to the current estimate, six morphotypes are found with unique taxonomic positions.The Anomochloatype and Streptochaeta-type were only found in the ancient Anomochlooideae subfamily.In Bambusoideae, the Diandrolyra-type, Olyra bahiensis-type, Olyra juruana-type, and Sucreatype were also unique (Fig. 10).It is worth emphasizing that a detailed sample analysis that incorporates fossils to estimate the changes in these morphotypes through time holds great promise.Such studies could indicate when the different grass pollen morphotypes evolved, and such knowledge could be used to measure changes in the diversity and composition of grasses through geological time.
As we are working with extant taxa, we cannot be sure when the morphotypes emerged.In principle, it is possible that they all were generated in the Cretaceous, but it is equally possible that they appeared in the Quaternary.The emergence of the variation in morphotypes could be tested by studying the fossil record through further palynological analysis.Some of our morphotypes are unique and potentially can be used to identify subfamilies.For subfamilies with a large variation in morphotypes, it will be difficult to use pollen morphotypes as an identifier.
In summary, we reported the current estimate for grass pollen surface ornamentation morphotype variation throughout their phylogeny.The pollen ornamentation morphotypes can be supported by quantitative analysis, and the degree of variation is consistent with the species richness of the subfamily.However, we are still limited as many Poaceae species are underexplored from a palynological point of view-the known pollen surface ornamentation of Poaceae is only 1.9% (223/11 783) (Soreng et al., 2022) of the total number of species.More work will be needed, on both extant and fossil grass pollen, and it is likely that many more morphotypes will be found in the future.We especially encourage scientists to focus on the surface ornamentation of fossil pollen to assess Poaceae surface variation types across the geological timescale.Further study is also needed to define if, and when, these morphotypes first occurred in the fossil record, and if the emergence of a large amount of diversity of pollen morphotypes occurred in the Cretaceous, when the Poaceae first evolved.Or if perhaps it was much later, in the Miocene, during the diversification of the grasses and the rise of grassland biomes (Palazzesi et al., 2022).

Implications for the wider research community
Linking the grass pollen morphotypes to databases detailing biogeographic, climatic, (paleo)botanical, and (paleo)ecological studies is already possible, permitting the application of tools developed for spatial and time ecology.Furthermore, paleoecologists would benefit from incorporating morphotypes to fossil grass pollen data, to elucidate the changes of these morphotypes through time.In the field of allergy and immunology studies, grass pollen is also utilized as a potential tool for diagnosing and/or preventing seasonal allergies.(i.e., Vinckier & Smets, 2001;Yasniuk et al., 2020) Our results may also be relevant in the context of allergy studies, as grass pollen morphology could play a role in allergic diseases.

Conclusion
We have examined the pollen from extant plant samples across northern South America to construct a reference framework for Poaceae pollen based on surface orna-mentation.The human-eye-level identified pollen morphotypes could be well supported by the extraction of two quantitative features (sculptural elements size and density) using computational analysis.The influence of vegetation type, soil type, photosynthetic pathway, elevation, and climate data does not seem to control the expression of a morphotype.The morphological richness of pollen surface ornamentation correlates with species richness at the subfamily level.Here, we represent an updated database of pollen surface ornamentation morphotypes of Poaceae, nevertheless, there is still some limitation as many Poaceae species remain unexplored (98.1% of the total amount of species).We encourage more scientists to focus on the combination of highresolution imaging of extant and fossil grass pollen surface ornamentation and computational image analysis to elucidate the (paleo-)ecology of grasslands.To validate if our finding based on modern material holds through evolutionary time next steps for this research would be to seek to identify the different morphotypes within pollen extracted from the fossil record.The (in)dependence of morphotype development from species evolution could provide valuable new insights into the mechanisms behind diversification.

Fig. 1 .
Fig. 1.Study area and localities where the herbarium specimens were collected and which we used in this study.Symbol colors denote the Poaceae subfamily; the map was modified from Wei et al. (2023).

Fig. 2 .
Fig. 2. Thumbnails showing the example of image processing steps that were taken during the construction of features size and density.A1-A6, Poa horridula Pilg.B1-B6, Tristachya leiostachya Nee.(A1-B1) shows raw images (1855 × 1855 pixels) with different morphotypes.(A2-B2) shows adjusting the image contrast and an example of randomly cropping subimages with 1000 × 1000 pixels.(A3-B3) shows the Sobel edge detection for each subimage.(A4-B4) shows the size feature for one of the connected components, with 1807 and 155 for P. horridula and T. leiostachya, respectively.(A5-B5) shows the four colors quantized for each subimage.(A6-B6) shows connected components using the brightest color as the foreground, which is used for counting the density feature for each subimage, with 60 and 21 for P. horridula and T. leiostachya, respectively.

Fig
Fig. 5.An example of Dendrocalamus asper showing the rugulate structure in grass pollen.Image credits: Priscila Fernandes de Souza, Pardinho, Brazil.

Fig. 6 .
Fig.6.Bar plot showing the frequency of each morphotype in terms of the number of plant specimens in our study.The light gray color is data from our sample, the dark gray color is data from published scanning electron microscope images from the literature.

Fig. 8 .
Fig. 8. Log size of Poaceae pollen ornamentation plotted against the six explanatory variables used in this study.A-C.Box plots of pollen ornamentation size against (A) vegetation type, (B) soil type, and (C) photosynthetic pathway.D-F, Scatter plots of pollen ornamentation size against (D) elevation, (E) mean annual temperature (°C), and (F) total annual precipitation (mm).The red number above the boxes represents the number of samples.The red line of the scatterplot represents a linear regression.

Fig. 9 .
Fig. 9. Log density of Poaceae pollen ornamentation plotted against the six explanatory variables used in this study.A-C, Box plots of pollen ornamentation density against (A) vegetation type, (B) soil type, and (C) photosynthetic pathway.D-F, Scatter plots of pollen ornamentation density against (D) elevation, (E) mean annual temperature, and (F) total annual precipitation.The red number above the boxes represents the number of samples.The red line of the scatterplot represents a linear regression.

Fig. 10 .
Fig. 10.Pollen surface ornamentation shows diversity across the Poaceae phylogeny.A Poaceae time tree estimated from molecular data and calibrated by fossil records, redrawn from Huang et al. (2022).BOP, Bambusoideae-Oryzoideae-Pooideae; PACMAD, Panicoideae-Arundinoideae-Chloridoideae-Micrairoideae-Aristidoideae-Danthonioideae.Numbers in brackets represent the species richness (species number with scanning electron microscope image/number of total species) among the subfamilies (Soreng et al., 2022).The pie chart represents the proportion of photosynthetic pathways of species in each subfamily (Appendix S2).The different colors and the number of symbols (square or circle) in the grid show the distribution of the 15-grass pollen morphotypes identified by each specimen in each subfamily.

Table 1
The description of 10 grass pollen surface morphotypes from our samples.

Table 3
Multitypes among specimens within species from existing publications.