Cell walls dominate the cellular allocation of silicon in sedges

Abstract

The degree of plant silicification is classically proxied by organ-scale silicon concentration, [Si], which ignores where Si is allocated at the cellular scale. Bridging this scale gap is crucial to elucidate the physiological and ecological roles of silicification in high-Si taxa and its influence on biogeochemical cycles. We used high-resolution X-ray microtomography and cellular-scale element mapping to study silica allocation and silicification patterns in 15 species from the second most-abundant Si-accumulating family worldwide, namely Cyperaceae (sedges). Despite considerable variation in deposition sites and cellular [Si] between species and anatomical areas, we show that at least 78±17% of silica is localized in the epidermis, mostly as a silica layer in the outer tangential cell wall (OTW). Both the thickness and cellular-scale [Si] of this continuous layer, which is up to 18 µm thick, were highly correlated to organ-scale [Si] among species. Our findings demonstrate that bulk silica content is largely governed by deposition in the OTW in sedges, rather than by deposition in cell lumina. These findings open new perspectives for Si-related plant physiological and biogeochemical processes and pave the way for the use of X-ray microtomography not only to visualize, but also to quantify biogenic silica allocation in plants.

Cyperaceae preferentially allocate silica to their cell walls where it forms a layer in the outer tangential wall of their epidermis, and whole-leaf Si can provide a proxy measurement for this layer.

Introduction

Plant silicification, the deposition of hydrated amorphous silica (nH₂O.SiO₂) in and between cells from soil-derived dissolved silicon (Si), plays a key role not only in enhancing plant resistance to abiotic and biotic stressors, but also in shaping trophic interactions and driving major biogeochemical cycles (Epstein, 1994; Conley and Carey, 2015; Cooke et al., 2016; de Tombeur et al., 2023). Over geological timescales, the interplay between plant-driven silicate weathering and biogenic silica in soils has significantly influenced both the global Si and carbon cycles (Conley and Carey, 2015). Within plants, the protective benefits of Si against biotic and abiotic stress are thought to derive from silica deposits rather than through direct biochemical activity (Coskun et al., 2019). This highlights the critical importance of silicification per se (Exley and Guerriero, 2019). Beyond its functional role in living plants, biogenic silica—preserved in soils and sediments—serves as a key proxy in paleoenvironmental reconstructions. As such, it provides insights on past climate, vegetation patterns and ecosystem dynamics (Strömberg et al., 2018).

Over the last decade, major advances have been made in deciphering in planta mechanisms underlying silicification (Kumar et al., 2017a, 2017b, 2021; Guerriero et al., 2018; Zexer and Elbaum, 2020; Zexer et al., 2023). In particular, it is increasingly accepted that some cell wall polymers such as hemicellulose and callose can act as templates for silica deposition by binding silicic acid and promoting its condensation as biogenic silica (Zexer et al., 2023). In addition, recent evidence in rice suggests the existence of genes controlling cell-specific silica deposition in leaves (Mitani-Ueno et al., 2023; Huang et al., 2025). The existence of such genes could explain the in planta redirection of soluble silicic acid from undamaged to manually damaged plant parts with subsequent silica deposition, as recently reported for wheat (Thorne et al., 2023). These discoveries have deepened our understanding of silicification processes in model crop species, such as rice, sorghum, and wheat. In contrast to the growing knowledge on silicification-related mechanisms, much less attention has been paid to how silica is spatially allocated within plant tissues (Hodson, 2019; Guerriero et al., 2020; de Tombeur et al., 2025).

Silica occurs in diverse forms—ranging from diffuse wall impregnation to discrete bodies in cell lumina (defined here as the space or cavity enclosed by a cell wall)—and exhibits remarkable variation in size, shape, and location (Kumar et al., 2017b; Hodson, 2019; Guerriero et al., 2020). Understanding where silica is deposited in different species is essential, as specific functions and trade-offs (e.g. defense against herbivores, physical strength, trade-offs with phenolics) might depend on the type and localization of silica deposits (Hodson and Guppy, 2022; de Tombeur et al., 2025). For instance, silica deposition in epidermal cell walls beneath the cuticle is thought to block fungal and herbivore penetration in rice—a phenomenon known as the ‘physical barrier hypothesis’ (Debona et al., 2017)—and probably increase leaf straightness (Yu et al., 2020). By contrast, the potential functions of silica deposits in cell lumina (e.g. silica cells in grasses) are unknown (Agarie et al., 1996), although the deposition process is tightly regulated and thoroughly studied (Kumar et al., 2017a; Kumar and Elbaum, 2018). In addition, whether silica is deposited in cell walls or lumina can have major consequences on biogeochemical cycles because they could have different dissolution rates once they enter the soil, as well as different C content trapped inside (Hodson, 2019). For instance, cell wall types where silica is deposited on a carbohydrate matrix have much greater C concentrations compared to those deposited inside the cell lumen (Hodson, 2019), which could have consequences on ecosystem-scale C cycling (de Tombeur et al., 2024). Yet, probably because of their use in paleoecology and archeology, deposits in cell lumina have received more attention than those in cell walls (Hodson, 2019).

The quantification of silica allocation in plant tissues is important. Proxying the degree of plant silicification by measuring bulk [Si] in leaves, as currently done (Hodson et al., 2005; Cooke and Leishman, 2011; Strömberg et al., 2016; de Tombeur et al., 2023; Pang et al., 2025), is not sufficient to fully appreciate the functional role of plant silicification and the consequences it has on ecological and biogeochemical processes. In particular, the allocation of silica to cell walls versus lumina is a key facet of silicification and related processes, which as yet remains poorly understood, especially in non-grass species (Hodson, 2019; de Tombeur et al., 2024). To address this challenge, we studied cellular-scale silicification at high-resolution in 15 species of Cyperaceae (sedges). Sedges were selected as they belong to the Poales—the highest Si-accumulating order among the angiosperms (Hodson et al., 2005)—and because they are well-represented globally and known to play a major role in Si biogeochemistry (Elliott et al., 2024). Despite this evidence, silicification in sedges remains understudied compared to grasses, even though they are the second most-abundant family within the Poales (Elliott et al., 2024). To visualize and quantify silica allocation in the studied species, two main techniques were used, X-ray microtomography (µCT), and cryo-SEM–energy dispersive X-ray spectroscopy (cryoSEM-EDX). µCT allowed visualization of biogenic silica in 3-D and quantification of its allocation in different anatomical areas, whilst cryoSEM-EDX allowed us to determine cell-specific [Si] in leaf or culm cross-sections.

Material and methods

Study area and species selection

The sampling was conducted in Alison Baird Reserve, a 35 ha flora reserve located 20 km south-east of Perth, in south-western Australia (32.020° S, 115.981° E). The reserve is part of the Swan Coastal Plain—a series of dunes parallel to the coast—and is located in a global biodiversity hotspot (Myers et al., 2000). The climate is Mediterranean, with hot, dry summers and cool, moist winters. The landscape in Alison Baird Reserve is variable, with low-lying areas to the east and west of a dune that can experience seasonal waterlogging, and drier high-lying areas formed by dunes, one of which runs through the middle of the reserve (Smith et al., 2023). Overall, soils across the reserve are very low in available P, sandy, highly weathered, and quartz-rich (Smith et al., 2023); soils typically have a low Si availability (de Tombeur et al., 2020a).

Fifteen perennial Cyperaceae from nine genera were selected in this reserve; nearly all of them being endemic to Western Australia: Chaetospora subbulbosa, Chorizandra enodis, Cyathochaeta avenaceae, Cyathochaeta equitans, Gahnia trifida, Lepidosperma leptostachyum, Lepidosperma pubisquameum, Lepidosperma rostratum, Machaerina junceae, Mesomelaena pseudostygia, Mesomelaena tetragona, Morelotia octandra, Schoenus caespititius, Schoenus rigens, and Schoenus subfascicularis.

A first sampling campaign was organized over a 2 d period in July 2022 in order to measure leaf thickness (or culm diameter), dry matter content (LDMC), mass per area (LMA), and the bulk [Si] and [C] among the 15 species. This sampling campaign is described in detail by de Tombeur et al. (2025), in which the LDMC, LMA, Si, and C results are published, and they are re-used here. Briefly, five individuals per species were sampled except for one rare species (Chorizandra enodis) for which only three replicates were found, making a total of 73 individuals. The main green, photosynthetically active organ was sampled (leaf or culm, depending on the species). For simplicity’s sake, ‘leaf’ is used when referring to a group of species in which some have leaves and some have culms, but the correct organ name (‘leaf’ or ‘culm’) is used when referring to a single species.

A second sampling campaign was conducted 2 weeks after the first campaign to sample three extra individuals of the same 15 species to conduct microscopic analyses via X-ray microtomography (µCT) scans and cryo-scanning electron microscopy energy dispersive X-ray spectroscopy (cryoSEM-EDX) analyses. The three individuals were sampled at the same location as those from the first sampling campaign, but were not necessarily the same individuals. For each individual, a 5 cm-long segment of leaf or culm was sampled.

Leaf morphological traits

For the first sampling campaign, we measured the leaf thickness (for species with leaves) or culm diameter (for species with culms) with a digital caliper. Briefly, LDMC was calculated as the ratio between leaf dry weight (after drying at 60 °C for 72 h) and leaf fresh weight, and LMA was calculated as the ratio between leaf dry weight and leaf surface area (scanned at 300 dpi and measured using ImageJ). For culms or round leaves, the projected surface was multiplied by π/2 before LMA determination, to take into account the greater light interception allowed by such shapes (de Tombeur et al., 2025). For species with leaves, we calculated the leaf fresh tissue density (ρ, g cm⁻³) as follows (Vile et al., 2005):

$$\rho ={\left(\mathrm{thickness}\times \frac{1}{\mathrm{LMA}}\times \mathrm{LDMC}\right)}^{-1}$$

For species with culms, we used the culm diameter (D) and total culm length (h; as scanned after rehydration) to calculate the total culm fresh volume (V) that was scanned, assuming that the culms were cylindrical: V=πh(D/2)². We then divided the culm fresh mass by the total culm volume that was scanned to obtain the culm fresh density, ρ (g cm⁻³). Results are shown in Supplementary Table S1.

Leaf Si, C, and phytolith concentrations

The leaf [Si] and [C] methodologies and results for the first sampling campaign are given by de Tombeur et al. (2025), and the results are re-used in this article. Briefly, we used a Li-metaborate and Li-tetraborate dissolution at 1000 °C to solubilize biogenic Si, and quantified it using an inductively coupled plasma optical emission spectrometer (ICP-OES). For C, we used the Dumas combustion method (vario MACRO cube, Elementar Analysensysteme GmbH, Langenselbold, Germany). The species studied showed a wide variation in their silicon concentrations, with species means ranging from [Si]=4.3–67.3 g Si kg⁻¹, and in their C:Si ratios ranging from ∼6–105 (Supplementary Table S2).

We then converted the leaf-level [Si] obtained after the chemical dissolution (DW basis) into a volume-based phytolith concentration (FW basis), assuming all Si was in the form of silica-based phytoliths. To do so, we first converted Si into phytoliths by multiplying [Si] by 2.14 (conversion factor from Si to SiO₂) and by 1.15 to add 15% by weight of water, C, and other elements commonly found in phytoliths and adding extra mass (Blecker et al., 2006). We then used the LDMC to express phytolith concentrations on a fresh mass basis. Finally, we used the leaf tissue density (as calculated above) and a phytolith density of 1.5 g cm⁻³ (in line with past studies; Aleman et al., 2013) to convert mass-based phytolith concentrations (g phytoliths kg⁻¹ FW) into volume-based concentrations (mm³ phytoliths cm⁻³ FW).

µCT scans

µCT analyses were conducted on material collected from the second sampling campaign to study leaf 3-D anatomy, including the location of potential mineral deposits of higher density compared with other tissues (biogenic silica; phytoliths). Analyses were conducted for all 15 species, each with three replicates. As contrast in µCT scans occurs as a consequence of differences in the relative X-ray attenuation of various anatomical structures due to their density composition, we expected high-density silica deposits to show up strongly in the data, and hence no contrast-enhancing stains were used. Thus, all image contrasts are the result of the inherent X-ray attenuation of the sample under the imaging conditions used.

The samples were mounted vertically (scanner y-axis) in 5 ml polypropylene tubes filled with 5% low-gelling agarose to ensure sample stability. The samples were scanned at 40 kV and 75µA using a VersaXRM 520 µCT system (Zeiss) running the Scout and Scan software (v.10.6.2005.12038, Zeiss). A total of 1600 projections were collected over 360°, each with a 5 s exposure. Binning (2×) was used to achieve a suitable signal-to-noise ratio and 4× optical magnification was used to achieve an isotropic voxel resolution of 3.8 µm. No source filters were applied. Projections were reconstructed using the XMReconstructor software (v.10.7.3679.13921, Zeiss) with a standard center shift and beam hardening correction, and a 0.7 kernel size reconstruction filter setting.

Data generated from µCT scans were analysed and visualized with Avizo 3D 2023.2 software. X-ray CT data were semi-quantitatively analysed to measure and visualize the phytolith concentration and distribution between species. Mean gray values were derived across multiple slices for air (spaces inside leaves/culms), agarose (surrounding the sample), and phytoliths in one species (M. tetragona). Mean air and agarose values were then determined for all other species. A linear approach comparing the slope and intercept of air and agarose against the known M. tetragona phytolith gray values was then used to determine a threshold value for phytoliths in the other species. These threshold values were used to identify phytolith voxels for volume analysis and visualizations.

We then represented the ‘average intensity projection’ for 50 consecutive slices of 3.8 µm each (voxel dimension) from a randomly selected region along the culms/leaves, which is equivalent to a 190 µm thick culm or leaf segment. We applied the ‘seismic’ color look-up-table that visualizes voxel grayscale values as black to blue, to white, then to red, with black and red voxels indicating weak and strong brightness, respectively. Air spaces appeared black (for instance inside mesophyll tissue). Parenchyma cells appeared dark blue, while sclerenchyma tissues appeared lighter blue than parenchyma and mesophyll cells, indicating higher density. Red voxels indicated areas of much higher density, such as mineral deposits (i.e. biogenic silica).

For all individuals, we calculated the volume fraction occupied by biogenic silica (voxels above the threshold value calculated for each species, using the linear method described above) as mm³ phytoliths cm⁻³ fresh leaf. These results were then compared with the ones calculated from the leaf-level leaf [Si] (after chemical digestion), leaf and phytolith density, and LDMC (see above). This comparison allowed us to test whether the two techniques of organ-scale [Si] after complete dissolution and µCT-derived concentration of phytoliths gave similar results. In addition to the fraction occupied by biogenic silica, we measured the thickness of the epidermal silica layer.

We next performed some segmentations to estimate the location of biogenic silica in different anatomical areas. To do so, within the fraction occupied by biogenic silica (only red voxels), we calculated the number of silica voxels (and therefore the volume in mm³) found in each of: the epidermal silica layer, conical phytoliths, surface siliceous prickle hairs, silica filling parenchyma cells, and silica found in the lumen of sclerenchyma cells. The results were then expressed as a percentage of the total biogenic silica in the samples. To visualize this segmentation process and where silica was located, we represented silica deposits in the different anatomical zones using different colors for a few selected species that had contrasting morphology and/or silicification patterns (Supplementary Fig. S1).

CryoSEM-EDX analyses

We performed cryoSEM-EDX analyses on material collected from the second sampling campaign to determine cell-specific [Si] and [C], using the same field samples as for the µCT analyses. In the field, culm/leaf segments were placed on an aluminum grooved pin with optimal cutting temperature compound and immediately plunge-frozen into liquid N, thus immobilizing and preserving cellular ions (Kotula et al., 2021; Oi et al., 2022). Samples were stored in liquid N until required.

We imaged and analysed the leaf/culm surfaces of one randomly selected replicate of each species. Data are shown only for a few selected species that exhibited specific Si hotspots on the leaf or culm surfaces, for example as defensive attributes, such as silicified prickle hairs in Gahnia trifida and in epidermal cell walls in Schoenus rigens.

For cross-section analyses, we were only able to examine nine out of the 15 species due to time constraints (Supplementary Table S3), and they were selected to maximize the variation in leaf [Si] and subsequent Si:C ratios. For these species, one cross-section was analysed for each of the three replicates. Two samples were lost during the storage process, resulting in a total of 25 cross-sections being analysed.

Perfectly flat transverse surfaces of frozen–hydrated culm/leaf samples were prepared by stepwise cryo-planing (1 μm; 0.5 μm; 0.25 μm; 0.1 μm) with a glass knife on a cryo-microtome (FC7, Leica Microsystems). Samples were transported using a transfer shuttle directly from the cryo-microtome to the cryo-preparation system (ACE600, Leica Microsystems), where they were coated with 20 nm chromium, without sublimation. Coated samples were then transported under vacuum in the transfer shuttle to a field emission SEM (JSM-IT800HR, JEOL) fitted with a cryo-stage (VCT500, Leica Microsystems) and an Oxford Ultim-Max170 X-ray detector interfaced to Oxford Instruments AZtec software. Samples were analysed at −155 °C, 15 kV, and a beam current of 1 nA (measured using a probe current detector in the column). Prior to sample analysis, the instrument was calibrated using a pure copper standard. Elemental maps were acquired at 512 pixel resolution, for >2000 frames with a dwell time of 10 μs per pixel. Drift correction and pulse-pile up correction were activated. For each of the 25 samples 3–7 quantitative fields of view (maps) were realized at different magnification (from ×190 to ×2500), for a total of 97 maps (Supplementary Table S3).

Quantitative numerical data were extracted from regions of interest (ROI) drawn on the maps, with individual spectra from each pixel summed and processed to yield concentration data using the Oxford Instruments AZtec software with inbuilt standards (Marshall, 2017; Kotula et al., 2021; Oi et al., 2022). Five ROIs were targeted: (1) the conical phytoliths found in epidermal cells (that are cell wall-type phytoliths; Mehra and Sharma, 1965; de Tombeur et al., 2024); (2) the outer tangential cell wall (OTW) of epidermal cells; (3) mesophyll cells; (4) parenchyma cells; and (5) sclerenchyma cells. The number of analyses for each ROI was not the same among the nine species (Supplementary Table S3), as it depended on the field of view and magnification of each mapped area. All the generated data are available in Supplementary Dataset S1.

Statistical analyses

We tested the associations between the theoretical volume-based phytolith concentrations calculated from organ-scale morphological and chemical data and those obtained through µCT through a standardized major axis (SMA) regression (Warton et al., 2006), using the package SMATR (Warton et al., 2012). The relationship between the thickness of the silica layer (obtained through µCT analyses) and the leaf-level [Si] was also tested through SMA regression.

After the segmentation performed on the µCT results, we tested if more or less silica was located in any given anatomical area for each species individually. To do so, we tested differences in the percentage of silica allocated to the different anatomical areas by ANOVA followed by Tukey’s post hoc test using the multcomp package in R (Hothorn et al., 2008).

For each of the nine species analysed by cryoSEM-EDX, the differences in cellular-scale [Si] and [C] between the different anatomical areas were also tested through ANOVA followed by Tukey’s post hoc test. Finally, among the same nine species, we used SMA regressions to test the associations between the [Si] of epidermal cell walls and the leaf-level [Si], and the [Si] and [C] of epidermal cell walls.

Results

To test the accuracy of µCT in quantifying the overall biogenic silica concentrations in plants, we compared the volumetric silica concentrations obtained from µCT (mm³ cm⁻³) with a theoretical estimate derived from organ-scale morphological and chemical data, namely leaf [Si], leaf density, and dry matter content (Supplementary Table S1). The two estimations of volume-based silica concentration were tightly correlated, with a slope close to 1 and an intercept close to 0, especially when the outlier Schoenus rigens was removed (Fig. 1A).

Fig. 1.

Silica allocation to different anatomical areas among the 15 studied sedges. (A) Standardized major axis (SMA) regression between volume-based silica concentration calculated either from leaf [Si], tissue density, and dry matter content (n=5) or from the µCT analyses (n=3). The regression line includes the outlier Schoenus rigens, and the y=x line is shown. The regression statistics are given with or without S. rigens, and the confidence intervals are shown for the slopes and intercepts. (B) SMA regression line of the relationship between the thickness of the epidermal silica layer (n=3) and the leaf-level leaf [Si] (n=5). All data are means (±SE). (C) Percentage allocation of biogenic silica to different anatomical areas, as determined in each biological replicate in each species (n=3); red dots indicate the mean values. Overall results of ANOVA conducted between the different anatomical areas are given: ***P<0.001; *P<0.05; ns, not significant. Different letters indicate significant differences between areas as determined by ANOVA followed by Tukey’s test (*P<0.05).

Segmentation analysis revealed that, in most of the studied species, silica was predominantly located in the outer tangential cell wall (OTW) of epidermal cells (Fig. 1C), forming a continuous layer just beneath the cuticle. The fraction of silica allocated to this region exceeded 50% for 12 out of the 15 species, and reached more than 80% for five species (Fig. 1C). The thickness of the OTW silica layer was positively correlated with organ-scale [Si] (Fig. 1B), highlighting the dominant contribution of this specific deposition site to overall silica accumulation. µCT cross-sections further confirmed that the epidermal cell wall was the primary site of silica deposition across all species examined, forming a continuous, shield-like layer around the leaf or culm (red signal in Fig. 2; Videos 1–6). CryoSEM-EDX corroborated the µCT results, revealing a distinct Si-rich layer in the OTW, immediately beneath the cuticle in the majority of the species analysed (Fig. 3A–E). Three species exhibited significant intraspecific variation in their silicification patterns, namely Lepidosperma rostratum (Fig. 2D), Machaerina junceae (Fig. 2G), and Schoenus rigens (Fig. 2L); µCT images of the two other replicates can be found in Supplementary Fig. S2.

Fig. 2.

X-ray microtomography images showing tissue density and silica deposit sites in the 15 studied sedges. A color look-up-table was applied to differentiate low density/background (black) through to higher-density tissues and mineral particles (blue to white to red). Each image shows an average projection image of 50 transverse µCT slices (totaling 190 µm thickness) for one representative replicate of each species. Air spaces appear in black, for example in the mesophyll region (green arrow in A). Water-filled parenchyma cells appear in dark blue (star in A), while sclerenchyma appear in lighter blue (orange arrow in A), indicating higher density. Red zones indicate areas of much higher density, such as mineral deposits. (A) Schoenus caespititius, (B) Chorizandra enodis, (C) Cyathochaeta avenaceae, (D) Lepidosperma rostratum, (E) Mesomelaena pseudostygia, (F) Cyathochaeta equitans, (G) Machaerina junceae, (H) Chaetospora subbulbosa, (I) Mesomelaena tetragona, (J) Schoenus subfascicularis, (K) Lepidosperma pubisquameum (L) Schoenus rigens, (M) Lepidosperma leptostachyum, (N) Gahnia trifida, and (O) Morelotia octandra. All scale bars are 300 µm.

Fig. 3.

Cell-specific silicon concentrations in selected individual sedges acquired using CryoSEM-EDX. For each species, a schematic diagram is shown of the different anatomical areas, together with a qualitative representation of the distributions of silicon (green) and carbon (purple) (Si-C), and a heatmap of silicon concentration, [Si]. (A) Machaerina junceae, (B, C, E) Cyathochaeta equitans, (D) Schoenus rigens, and (F) Mesomelaena tetragona. Scale bars are 100 µm in (A–D, F) and 25 µm in (E).

Silica was also prominently present as conical phytoliths embedded in the cell walls of epidermal cells (Fig. 1C), further contributing to the overall silicification pattern. The fraction of silica allocated to these phytoliths ranged from none to ∼50%. Conical-shaped phytoliths were arranged in rows immediately below the leaf or culm surface (Figs. 2A, 4A; Video 2) and exhibited high [Si] (Fig. 3E). For two very abrasive species, ∼25% of all silica was allocated to large and sharp prickle hairs on the leaf surface (Fig. 1C). Three-dimensional µCT reconstructions and cryoSEM-EDX Si maps of these silicified, abrasive prickle hairs are shown in Fig. 4B C, and in Video 1. When summed across all epidermal structures—OTWs, conical phytoliths, and prickle hairs—the proportion of silica allocated to the epidermis ranged from 78±17% to 100±10% across all 15 species.

Fig. 4.

3-D volumes obtained by X-ray µCT and CryoSEM-EDX mapping showing conical-shaped phytoliths, silicified prickle hairs, and silicified parenchyma cell lumina. (A) Rows of conical-shaped phytoliths in Schoenus caespititius shown with or without the rest of the culm; scale bars are 200 µm. A magnified image of one row is shown below (scale bar is 50 µm). (B) Silicified prickle hairs on the leaf surface of Gahnia trifida. The middle image shows only the hairs (densest areas of the whole leaf). Scale bars are 200 µm. (C) Back-scattered electron microscopy image of surface prickle hairs in G. trifida (left), together with a qualitative representation of the distributions of silicon (green) and carbon (purple) (middle), and a heatmap of silicon concentration, [Si] (right). Scale bar is 50 µm. (D) Denser zones in Cyathochaeta avenaceae showing deposits in prickle hairs (large deposits on edges and occasionally on surface), as rows of conical-shaped phytoliths (smaller deposits in well-arranged in rows), in epidermal cell walls (diffuse layer), and filling internal parenchyma cells; scale bar is 200 µm. A magnified image of the parenchyma cells is shown on the right (scale bar is 70 µm). All µCT images represent 250 slices of a culm/leaf segment ∼950 µm long.

In addition to the silica deposits in the epidermal region, we found silica occasionally filling the lumen of internal parenchyma cells in some species (Fig. 1C). A 3-D example of such deposits is given in Fig. 4D. CryoSEM-EDX further demonstrated the existence of such deposits by showing very high [Si] in some cases (∼400 g kg⁻¹; Fig. 5A). Silica was also found filling the lumen of sclerenchyma cells in certain species (Fig. 6A; Videos 5, 6). These deposits appeared as elongated columns running along the length of the culm, sometimes reaching substantial dimensions of up to ∼1 mm. CryoSEM-EDX analyses confirmed that, in some instances, Si in sclerenchyma cell lumina reached remarkably high concentrations (Fig. 5B). Finally, the species with the highest [Si] (Schoenus rigens) exhibited distinctive silicification patterns, especially in one individual (Fig. 2L, 3D, 6B–D; Video 4) where it was found as unusual large deposits within the epidermal cells, as well as in epidermal cell walls that were fully silicified.

Fig. 5.

Cell-specific silicon concentrations in selected individual sedges acquired using CryoSEM-EDX. (A, B) Schematic diagrams of the different anatomical areas, together with qualitative representations of the distributions of silicon (green) and carbon (purple) (Si-C), and heatmaps of silicon concentration, [Si] for (A) Cyathochaeta equitans and (B) Mesomelaena tetragona. Scale bars are 100 µm. (C) Box-plots showing the concentration of Si in five different regions of interest. The line inside each box shows the median, the box represents the interquartile range, and the whiskers indicate the locations of the most extreme data points that were still within a range of 1.5 of the upper- or lower-quartiles; the dots represent all the individual measurements. The y-axis has been root-square-transformed to improve visualization. Overall results of ANOVA conducted between the different anatomical areas are given: ***P<0.001. Different letters indicate significant differences between areas, as determined by ANOVA followed by Tukey’s test (*P<0.05). Three biological replicates were analysed by cryoSEM-EDX for each species, except for M. tetragona and G. trifida where n=2 (see Supplementary Table S3). (D) Standardized major axis (SMA) regression between the leaf-level Si concentrations (n=5) and the mean Si concentrations in the epidermal cell walls for the nine species analysed by cryoSEM-EDX (n=3 biological replicates, but numerous epidermal cell wall analyses were made for each replicate; see Supplementary Table S3). (E) SMA regression between the mean Si and carbon concentrations in the epidermal cell walls for the nine species analysed by cryoSEM-EDX (n=3 biological replicates, but see Supplementary Table S3). Data are species means (±SE).

Fig. 6.

3-D volumes obtained by X-ray µCT and CryoSEM-EDX mapping showing long silica deposits in the sclerenchyma cell lumen and in epidermal cell walls. (A) Long silica deposits filling the lumen of sclerenchyma cells in Machaerina junceae. The two panels on the left show columns of silica based on 250 slices of a culm segment ∼950 µm long; scale bars are 200 µm. The on the right right is for 930 slices of the same individual (culm segment ∼3.5 mm long) showing the potential length of these sclerenchyma silica columns (in red); scale bar is 500 µm. (B) Gradient of silica density (increasing density from top to bottom) within the silica deposits in Schoenus rigens (250 slices; scale bar is 200 µm). Silica is mostly found as a thick layer in epidermal cell walls, but also filling large cells (see also Fig. 2L). (C) Silica filling the walls of the epidermal cells in another individual of S. rigens (250 slices; scale bar is 200 µm). (D) Back-scattered electron microscopy image of silicified cell walls of epidermal cells on the leaf surface of S. rigens (left), together with a qualitative representation of the distributions of silicon (green) and carbon (purple) (middle), and a heatmap of silicon concentration, [Si] (right). Scale bar is 50 µm.

Cell-level [Si] tended to decrease in the following order: conical phytoliths > epidermal cell wall > mesophyll ≈ parenchyma ≈ sclerenchyma (Fig. 5C). The [Si] of conical phytoliths generally ranged between ∼200 g Si kg⁻¹ and ∼300 g Si kg⁻¹, with maximal values around 340 g Si kg⁻¹. The [Si] in OTW was much more variable, with concentrations ranging from <1 g Si kg⁻¹ to >300 g Si kg⁻¹. Silicon was occasionally detected in mesophyll, parenchyma, and sclerenchyma cells, but mostly in much lower concentrations (often below 10 g Si kg⁻¹). However, the lumen of both parenchyma and sclerenchyma cells was also occasionally filled with silica, with the [Si] in these areas reaching high values as well (e.g. up to 340 g Si kg⁻¹ in parenchyma cells of C. equitans). Interestingly, the leaf-level [Si] was positively correlated with the [Si] in the OTW (Fig. 5D). This further showed the strong effect of Si accumulation in the epidermal silica layer on whole-organ [Si]. In addition, concentrations of Si and C in the epidermal cell walls were negatively correlated with each other (Fig. 5E).

Discussion

This study represents the first attempt to quantify plant silica allocation at the cellular scale, and enhances our understanding of plant silicification from a methodological and conceptual perspective. Overall, our results also showed that most of the silica was allocated to cell walls, and much less to cell lumina (Fig. 1C). In particular, in spite of widely diversified silica deposits, a consistent pattern of allocation to the epidermis was identified in all species. Most silica was found as a layer located in the outer tangential cell wall (OTW) of the epidermis (Fig. 2). The thickness of this layer and its [Si] correlated with the organ-scale [Si] between species (Figs 1B, 5D), meaning that it represents a relevant proxy for the overall Si accumulation at the whole-organ scale. From a methodological perspective, our results demonstrate that µCT results correspond very well with leaf Si concentrations measured at the organ-scale (Fig. 1A). The strong convergence between these two methods for measuring the degree of silicification also demonstrates that Si in plants is mostly found as biogenic SiO₂ that can be identified by µCT because of its density, and that all the mineral deposits in the studied sedges are silica-based.

The potential ecophysiological roles behind the wide variation in Si accumulation among terrestrial plants remains difficult to understand (de Tombeur et al., 2023). While fertilization experiments have led to many roles being assigned to Si (Cooke and Leishman, 2016; Debona et al., 2017), such as defense against herbivores and pathogens, and better resistance to numerous abiotic stresses, clear links with the natural variation in organ-scale leaf [Si] between species remain to be established (de Tombeur et al., 2023). In this context, ignoring how variation in organ-scale [Si] between species is reflected at the cellular-scale significantly limits our understanding of the potential functions of Si (de Tombeur et al., 2023). Here, we show that silica deposited as a layer in the OTW of the epidermis, as well the thickness of this layer, is a good predictor of organ-scale leaf [Si]. This implies that organ-scale [Si] can reasonably be used in comparative studies as a proxy for the thickness and extent of this layer, at least in perennial sedges. Such a silica layer most likely acts as a protecting sheath against different stresses. It could limit excessive water loss through evapotranspiration (Agarie et al., 1998) and increase resistance to herbivores and pathogens (Debona et al., 2017). Furthermore, silica deposits could also reduce the amount of UV radiation entering the tissue, as shown in leaves of okra (Pierantoni et al., 2017). Whether the formation of such a layer is a tightly controlled process involving C-based polymers (Zexer et al., 2023) or a passive process through Si-fluxes driven by evapotranspiration is unknown. Further investigation of the pathway for formation and regulation of such silica layers is of high importance, given its significant contribution to whole-organ [Si].

Our results also showed that silica can sometimes be found in sharp prickle hairs on leaf surfaces, as identified in at least two of the studied species (Figs 1C, 4B, C). Interestingly, these species were the only ones to have a particularly abrasive and sharp leaf. Although no quantifiable links can be established here, it is very likely that silica deposits in prickle hairs significantly reinforce leaf abrasiveness, with consequences for herbivore deterrence. Direct links between leaf surface abrasiveness and the quantity of prickle hairs have been established in grasses (Hartley et al., 2015). In contrast with such deposits for which a direct role seems very likely, the functions of silica found in the lumina of sclerenchyma cells, as observed in a few species in the form of elongated columns running along the length of the culm (Fig. 4D), are less clear. Recent evidence in date palm (Phoenix dactylifera) suggests that silica deposited in cells of the sclerenchyma bundles is tightly controlled by Si influx transporters (Bokor et al., 2019). This might suggest that these deposits have specific functions, but it remains to be tested. Since sclerenchyma tissues are mostly made of fibers and used for plant strength and support (Crang et al., 2018), silica inside lumina could further enhance their biomechanical properties (Bokor et al., 2019). Combining trait-based approaches and microscopical tools in comparative studies is valuable in better understanding the potential functions of silicification (de Tombeur et al., 2025; Kajino et al., 2025), and our present results demonstrate that µCT is an efficient technique for achieving this.

A clear and general pattern that emerges from our study is that most of the silica is found in cell walls, and much less in the lumina. Evidence is increasing that some C-based polymers, such as callose or hemicellulose, can initiate and even control the degree of silicification in cell walls (Zexer et al., 2023). For instance, callose deposition in the trichomes of Arabidopsis seems essential for cell wall silicification (Kulich et al., 2018). Such a tight linkage between C-based compounds and silica deposition raises the question of the costs of accumulating silica. While these costs have long been considered as very cheap compared with those of lignin or cellulose (Raven, 1983; but see Cooke and Carey, 2023; de Tombeur et al., 2023), the requirement of specific C polymers to control silicification would make them significantly higher, especially if most silica is deposited in cell walls. Nevertheless, C polymers synthesized specifically for the purpose of providing a matrix for silica deposition would need to be identified to confirm such a hypothesis. This is important because not being able to properly characterize the costs of Si accumulation has probably hampered our understanding of the evolution of silicification in plants (de Tombeur et al., 2023).

In addition to consequences on the roles and costs of Si, the large dominance of cell wall silica deposits could have implications for biogeochemical cycles. First, the stability of phytoliths when they enter the soil depends on their shape, surface area, and degree of silica condensation (Schaller et al., 2021). In particular, it is assumed that silica from cell lumen phytoliths could have a higher degree of condensation compared to cell wall phytoliths (Schaller et al., 2021), and therefore be more stable once they are in the soil (Hodson, 2019). This needs to be further investigated, but it would imply that biogenic silica from sedges is particularly labile, and contributes to the biological Si cycle (Alexandre et al., 1997; de Tombeur et al., 2020b) perhaps more strongly than taxa that have more lumen phytoliths. Second, the content of C trapped inside phytoliths (termed PhytOC) is strongly dependent on the phytolith type (Hodson, 2019; de Tombeur et al., 2024). Silica deposited on a carbohydrate matrix tends to have much higher C concentrations compared with lumen phytoliths, which may have consequences on ecosystem-scale C cycling (de Tombeur et al., 2024). Using our silica allocation results and assuming that phytoliths have C concentrations of 25% for the OTW, 15% for conical phytoliths, and 2.5% for cell lumen phytoliths (Hodson, 2019; de Tombeur et al., 2024), we calculate that the 15 sedges exhibit an overall PhytOC concentration of 22%, with a range of 17.6–24.4%. These values are much higher than those obtained with previous methods (Hodson, 2019), but cell-wall phytoliths dominated in the sedges studied and we might expect different results in grass leaves, where lumen phytoliths are more abundant.

Understanding why plants vary so much in their degree and pattern of silicification remains a fundamental and important question in plant biology (Epstein, 1994; Coskun et al., 2019; Exley and Guerriero, 2019; de Tombeur et al., 2023). Addressing it requires both interdisciplinary approaches and closer integration of ecological and anatomical perspectives. In particular, it is essential to reconcile ecological comparative approaches that mostly consider Si as a concentration at the organ level with microscope-based approaches that consider Si allocation patterns, but are mostly limited to a handful of model species. Here, we show that ∼80% of silica is found in the epidermis of sedges, primarily as a continuous silica layer in the OTW of the epidermal cells, the thickness of which directly explains the variation in organ-scale [Si] among species. Further studies need to be conducted on species belonging to other high-Si families (e.g. see Pierantoni et al., 2017), whether that be abundant ones such as Poaceae or scarcer ones such as Juncaceae or Restionaceae, in order to draw general conclusions regarding plant silicification patterns and the role of cell walls as major deposit sites. In addition, the species studied here have high Si concentrations (up to 67 g kg⁻¹) and well-defined silica deposits, which made them ideal to use µCT as a semi-quantification tool, but the accuracy and resolution of µCT to both visualize and quantify silica allocation in plants needs to be further verified with species with lower Si concentrations.

Supplementary data

The following supplementary data are available at JXB online.

Fig. S1. Examples of the segmentation performed on µCT 3-D volumes.

Fig. S2. X-ray µCT images highlighting the intraspecific variation in silicification patterns for three species.

Table S1. Leaf morphological properties among the 15 studied species.

Table S2. Leaf chemical properties and volume-based phytolith concentrations among the 15 studied species.

Table S3. Summary of the CryoSEM-EDX analysis conducted on nine species.

Dataset S1. Data generated by the CryoSEM-EDX analysis.

Author contributions

FdT, PLC, PEH, LP, CV, MJH, and HL planned and designed the research; LP, FdT, PLC, JK, JS, KR, and PH performed the experiments; FdT, MJH, RL, PLC, and JS analysed the data; FdT wrote the first version of the manuscript, with input from all the other authors.

Funding

3D data processing was performed using the µCT facilities of the Labex CEMEB, which receives funding from the French National Research Agency (ANR, project no. 10-LABX-0004). This project has received funding from the European Union’s ‘Horizon 2020’ research and innovation programme under the Marie Skłodowska-Curie grant agreement no. 101021641 (project SiliConomic) to FdT, and from the ANRfor the project ‘PhytoStone’ (project no. 25-CE02-3771) to FdT.

Data availability

All data supporting the findings of this study are available within the paper and within its supplementary materials published online.