Whipping phytoliths into shape (and size). A Commentary on: ‘Inter- and intraspecific variation in grass phytolith shape and size: a geometric morphometrics perspective’

Abstract

This article comments on:

Kristýna Hošková, Adéla Pokorná, Jiří Neustupa and Petr Pokorný, Inter- and intraspecific variation in grass phytolith shape and size: a geometric morphometrics perspective, Annals of Botany, Volume 127, Issue 2, 01 February 2021, Pages 191–201, https://doi.org/10.1093/aob/mcaa102

The study of phytoliths (Greek for ‘plant stones’) has, in the last 50 years, become standard practice among archaeobotanists and palaeoecologists for addressing questions as disparate as regarding plant evolution, palaeoenvironmental reconstruction and plant domestication. However, a problem that has long plagued phytolith science is the extent to which phytolith morphology can be used to distinguish between different plant groups. The question at the heart of this problem is: just how variable are phytoliths within and among taxa? The study by Hošková et al. (2020) addresses this question head on in its elegant and robust application of 2D geometric morphometrics to grass silica short cell phytoliths (GSSCPs) of two species in the genus Brachypodium.

Members of all major groups of land plants deposit phytoliths as microscopic, hydrated silica bodies in and around their cells. The inert composition and general hardiness allow phytoliths to be preserved in archaeological and geological settings, even when other plant fossils do not, which in part explains their utility for investigating plant communities and plant–human interactions on short and long timescales. Phytoliths also happen to be the most reliable source of fossil evidence for tracking the evolutionary history of grasses (Poaceae), arguably among the most important extant plant families. Over the last ~25 Myr, Poaceae expanded to cover >40 % of Earth’s terrestrial surface, and also make up the bulk of what humans eat. Whereas grass fossils such as pollen, leaves or diaspores are either rare or not taxonomically informative below the family level, grass phytoliths, especially the so-called GSSCPs, are known to vary in shape depending on grass taxon (subfamilies, tribes, sometimes even genera). Given sufficient contextual information, this shape variation allows grass taxa to be identified from fossil GSSCPs, a classic example being Zea mays phytoliths signalling the rise of agriculture in Mesoamerica (Piperno, 2006).

Nonetheless, meaningfully quantifying GSSCP morphology has proven difficult, in part because of the complexities in GSSCP production. The epidermis in most grass leaves contains two or more GSSCP shape-types (i.e. morphotypes) that form above (costally) or between (intercostally) veins. Many of these morphotypes are not unique to species but show varying degrees of morphological overlap with other taxa (Gallaher et al., 2020). Therefore, GSSCP assemblages, rather than individual GSSCPs, must be considered when attributing fossil phytoliths to existing taxa. Furthermore, several studies have demonstrated GSSCP shape/size variation between organs (e.g. leaves vs. inflorescences) (e.g. Novello and Barboni, 2015), different leaves (old vs. young), or even parts of leaves on individual plants. Finally, GSSCP shape/size variability among plants or plant populations has been noted and typically interpreted as plastic responses to environmental gradients (e.g. aridity, light) (e.g. Dunn et al., 2015). Because of these complexities, phytolith workers have traditionally focused on morphotypes that are unique or abundant in taxa or functional groups; their relative frequencies are used to reconstruct past grass community composition (Piperno, 2006).

However, even if polymorphism and redundancy in grass phytolith production is ignored, the fundamental problem of evaluating similarity or difference between individual GSSCPs remains. GSSCP morphotypes are typically defined based on qualitative or semi-quantitative descriptions of their 2D shape in planar view (e.g. number of lobes, whether elongated or rounded), although 3D criteria are occasionally included (e.g. Piperno and Pearsall, 1998). This approach is subjective and relies on ‘expert’ knowledge, resulting in classification schemes that are difficult to replicate or very coarse and include no means to assess the statistical robustness of phytolith identifications. The qualitative morphotype approach also does not accommodate intermediates between morphotypes well, further limiting its use. Recent studies using linear measurements, shape factors and outlines of GSSCP 2D shape (e.g. Liu et al., 2016) signify an improvement, but fail to fully capture GSSCP shapes and do not allow comparisons between different morphotypes.

In recent years, the call for objective and quantitative ways to measure phytolith shape has grown louder (Evett and Cuthrell, 2016), reflecting advances in imaging and data analysis that increasingly promote automation in plant morphology studies (e.g. Li et al., 2017). Phytolith workers have been late to seize on this opportunity, but just this year, studies tackling shape differentiation are beginning to emerge and are paving the way for future research. One of them, Gallaher et al. (2020), applies geometric morphometrics (GMM) to GSSCP 3D shape/size in 70 grass species in the early-diverging grasses, Oryzoideae, and Bambusoideae to assess broad phylogenetic patterns of morphological evolution and use these to classify Eocene (45–35 Mya) fossil GSSCPs. Their work demonstrates that GSSCP morphology carries a phylogenetic signal that can distinguish grass subfamilies and tribes (see Fig. 1), but leaves questions surrounding intra- vs. interspecific shape/size variation unanswered.

Fig. 1.

Levels of variation in grass silica short cell phytolith (GSSCP) shape and size. (A) Subclades and species in the Poaceae phylogeny. (B) Populations within a species. (C) Individual plants within a population, individual leaves and different parts of leaves (different distance to the basal intercalary meristem). (D) Costal vs. intercostal areas of leaves. (E) Among and between GSSCP morphotypes. In their paper, Hošková et al. (2020) investigate all levels in their study of two grass species in the genus Brachypodium.

Addressing core questions about fine-scale morphological variability is precisely the goal of Hošková et al. (2020). By applying GMM to 2400 GSSCPs from two Brachypodium species (B. pinnatum, B. sylvaticum), they are able to, for the first time, quantitatively and objectively assess the entire 2D phytolith shape, and to compare across several GSSCP morphotypes (as traditionally defined) in the same morphospace. Furthermore, for each species, the authors sampled multiple plants in multiple populations, and, for each plant, multiple leaves (old vs. young) and parts of leaves (apical vs. basal). This nested study design allows Hošková and colleagues to test several long-standing hypotheses regarding intra- vs. interspecific GSSCP variation. They show that nearly half of the observed shape and size variation in GSSCPs is linked to intra- and interspecific factors. However, whereas shape varies mostly (34 %) between species, size differences relate primarily (28 %) to intraspecific sources. This is good news for phytolith science, because it indicates that lineage, not ontogeny or environment, is the most important factor in shape variation among grass phytoliths. In other words, the use of phytoliths to infer taxonomic composition of past grass communities or identify ancient crop species is justified, at least for closely related species. Indirectly, these results also bolster those of Gallaher et al. (2020), looking at broader phylogenetic patterns. Hošková et al. (2020) further use their GSSCP shape data to classify phytoliths into either of their two species with 83 % accuracy. This approach importantly lays the foundation for assigning probabilities to taxonomic identifications in the fossil record (see also Gallaher et al., 2020).

Although Hošková et al. (2020) evaluate only two species, the protocol that they establish will be vital for continuing to test the role of inter- vs. intraspecific variation and creating a comprehensive database of GSSCP morphological data. Further testing is crucial, as previous work has pointed to lineage-specific differences in phenotypic plasticity in phytolith shape and size (Dunn et al., 2015). It will be particularly important to understand how size and shape vary along environmental gradients, whether this variation hampers taxonomic assignments, or if it could constitute a useful proxy for past environmental conditions. As the authors themselves note, their method requires some work to be widely applicable, and might initially serve as a complement to more traditional approaches. Nevertheless, the work by Hošková et al. (2020) signals a new era of quantitative, objective and repeatable practices that will vastly improve accuracy and robustness in phytolith-based archaeology and palaeoecology.