Formation of root silica aggregates in sorghum is an active process of the endodermis

Silica, a mineral reducing the digestibility of grass tissues, is assumed to form spontaneously in planta. We overturn this model, showing that silica deposition occurs during lignification and demands metabolic energy.

Abstract

Silica deposition in plants is a common phenomenon that correlates with plant tolerance to various stresses. Deposition occurs mostly in cell walls, but its mechanism is unclear. Here we show that metabolic processes control the formation of silica aggregates in roots of sorghum (Sorghum bicolor L.), a model plant for silicification. Silica formation was followed in intact roots and root segments of seedlings. Root segments were treated to enhance or suppress cell wall biosynthesis. The composition of endodermal cell walls was analysed by Raman microspectroscopy, scanning electron microscopy and energy-dispersive X-ray analysis. Our results were compared with in vitro reactions simulating lignin and silica polymerization. Silica aggregates formed only in live endodermal cells that were metabolically active. Silicic acid was deposited in vitro as silica onto freshly polymerized coniferyl alcohol, simulating G-lignin, but not onto coniferyl alcohol or ferulic acid monomers. Our results show that root silica aggregates form under tight regulation by endodermal cells, independently of the transpiration stream. We raise the hypothesis that the location and extent of silicification are primed by the chemistry and structure of polymerizing lignin as it cross-links to the wall.

Introduction

Amorphous hydrated silica (SiO₂·nH₂O, herein silica) is one of the most common biominerals in the plant kingdom (Bauer et al., 2011; He et al., 2014; Coskun et al., 2019). Its deposition in plant cell walls, cell lumen, and intercellular spaces modifies the properties of plant tissues and supports their resistance against various biotic and abiotic stresses (Currie and Perry, 2007; Yamanaka et al., 2009; Liu et al., 2013; He et al., 2015; Ma et al., 2015). The chemical form of Si available for plant absorption is mono-silicic acid, Si(OH)₄. Silicic acid is taken up by roots either passively or actively (Ma and Yamaji, 2015), distributed to target locations in the plant, and mineralized into silica. The silicified structures are often referred to as silica phytoliths (Bauer et al., 2011). Although silicification is a well-known phenomenon, the mineralization process and its regulation are unknown (Guerriero et al., 2016; Kumar et al., 2017b).

Many studies indicate interactions between the deposition of silica and lignin/phenolic compounds (Inanaga and Okasaka, 1995; Fang and Ma, 2006; Schoelynck et al., 2010; Fleck et al., 2011, 2015; Suzuki et al., 2012; Yamamoto et al., 2012; Zhang et al., 2013; Hinrichs et al., 2017; Klotzbücher et al., 2018). Furthermore, silica deposits are frequently associated with lignified tissues (Scurfield et al., 1974; Piperno et al., 2002; Zhang et al., 2013). However, a direct link between plant tissue silicification and lignification is missing.

In lignifying plant cell walls, a variety of phenolic compounds participate in radical coupling reactions and incorporate into the growing lignin polymer. The fundamental subunits to form lignin are coniferyl alcohol (CA), p-coumaryl alcohol, and sinapyl alcohol, producing the G-, H- and S-subunits of the lignin polymer, respectively. The proportions of the lignin subunits vary between species, with G-lignin as the most abundant in grasses (Barros et al., 2015). Ferulic acid (FA) is not one of the three canonical monolignols. Nonetheless, ferulates are a natural component of lignins in grasses and could be considered a monolignol as well (Ralph et al., 2008). In addition, FA cross-links the non-lignified cell walls via FA–arabinoxylan (AX) complexes.

In the roots of sorghum (Sorghum bicolor (L.) Moench; Poaceae), silica is deposited in large quantities as discrete silica aggregates, anchored within the inner tangential cell walls (ITCWs) of the root endodermis (Sangster and Parry, 1976a,b,c; Hodson and Sangster, 1989; Lux et al., 2002, 2003). Our previous work shows that the silica aggregates are associated with the thickened cell walls, which appear only after the ITCW lignification initiates, and contain traces of the FA-AX complexes (Soukup et al., 2017). These observations indicate a possible interconnection between the deposition of silica and phenolic compounds. Thus, in this study we aimed to elucidate the relationship between silica aggregate formation, cell wall deposition, and lignification in the root endodermal cell walls.

Material and methods

Plant material

Grains of Sorghum bicolor (L.) Moench, line BTx623, were surface-sterilized with 2.5% sodium hypochlorite for 15 min, washed with distilled water, and imbibed in distilled water for 24 h. Grains were then placed into wet paper rolls and germinated for 72 h.

Hydroponic cultivation (also referred to as in planta cultivation)

After germination, the seedlings were grown hydroponically for 6 d. The first 3 d they grew in distilled water supplied with either sodium silicate (Na₂O(SiO₂)_x·xH₂O) at final concentration 1 mM (Si+ medium) or NaCl at final concentration 1 mM (Si− medium) to preserve similar ionic balance of the media. Final pH of the solutions was adjusted to 5.8 with HCl. After 3 d of cultivation, the growth solutions were changed and the seedlings were grown for additional 3 d to obtain the following growth conditions: (i) seedlings grown in Si− medium throughout the entire cultivation (Si− treatment), (ii) seedlings grown in Si+ medium throughout the entire cultivation (Si+ treatment), (iii) seedling grown for 3 d in Si− medium and for 3 d in Si+ medium (Si−/Si+ treatment), or (iv) 3 d in Si+ medium and 3 d in Si− medium (Si+/Si− treatment). The cultivations were performed in a growth chamber, under controlled conditions with photoperiod 16 h–8 h (light–dark) illuminated with photosynthetically active radiation (PAR) of approximately 200 μmol m⁻² s⁻¹, 28 °C–22 °C (light–dark) temperature, and 70% air humidity. Root samples were then collected from primary root region 1–3.5 cm from the primary root–shoot junction.

Ex planta cultivation of detached primary root segments (based on Soukup et al., 2017)

Silica is a common contaminator of many salts, because of its omnipresence. To avoid exposure to silicic acid we chose to grow our samples in a minimal medium, taking into account the important nutritional supply of the seed. Therefore, after germination, seedlings were precultivated hydroponically in distilled water for 72 h under Si− conditions as described above for the hydroponic cultivations. Afterwards, seedlings of similar morphology, with emerged lateral roots and primary root length exceeding 6 cm were selected and used for root segment preparation. Segments were collected from the primary root region 1–3.5 cm from the root–shoot junction. Lateral roots emerging from the segments were cut off. Some segments were pretreated before the ex planta cultivation (Table 1). Prepared segments were placed in Erlenmeyer flasks containing 50 ml of the cultivation medium (Table 1). The pH of all cultivation media containing sodium silicate was adjusted to 5.8 before adding other components. The flasks were sealed with Parafilm and cultured for 72 h (unless stated otherwise) in the dark at 24 °C, with permanent shaking at 60 rpm. After cultivation, the segments were fixed in FAA solution (3.7% formaldehyde, 50% ethanol, 5% glacial acetic acid, 41.3% distilled water; v/v).

Table 1.

The pretreatments and treatments applied to detached sorghum root segments cultivated ex planta

Treatment ID	Segment pretreatment	Compositiona
Si+ 1 mMb		Sodium silicate 1 mmol dm−3
Si+ 2 mM		Sodium silicate 2 mmol dm−3
Si+ 2 mM sucrose		Sodium silicate 2 mmol dm−3, sucrose 1% (w/v)
Si−		NaCl 1 mmol dm−3
Removed cortex	Rhizodermis and cortical tissues peeled-off mechanically	Sodium silicate 1 mmol dm−3
Liquid nitrogen	Segments kept in liquid nitrogen for 5 min	Sodium silicate 1 mmol dm−3
DNPc		Sodium silicate 1 mmol dm−3, 2,4-dinitrophenol 0.01 or 0.05 mmol dm−3
Low temperaturec	Erlenmeyer flask with the segments cultivated in ice-filled Styrofoam box (temperature 2–4 °C)	Sodium silicate 1 mmol dm−3
Sucroseb		Sodium silicate 1 mmol dm−3, Sucrose 1% (w/v)
Arabinoseb		Sodium silicate 1 mmol dm−3, arabinose 1% (w/v)
ATP		Sodium silicate 1 mmol dm−3, ATP 1 or 5 mmol dm−3
Brefeldin Ab		Sodium silicate 1 mmol dm−3, 0.01% ethanol (v/v), brefeldin A 1, 5 or 10 µmol dm−3
SHAMb		Sodium silicate 1 mmol dm−3, salicylhydroxamic acid 0.2 or 2 mmol dm−3
KIb		Sodium silicate 1 mmol dm−3, potassium iodide 0.5 or 5 mmol dm−3
Ascorbic acidb		Sodium silicate 1 mmol dm−3, ascorbic acid 1 or 5 mmol dm−3
Ferulic acidb		Sodium silicate 1 mmol dm−3, ferulic acid 0.05 or 0.2 mmol dm−3
H2O2b		Sodium silicate 1 mmol dm−3, hydrogen peroxide 1 mmol dm−3
H2O2 pretreatment	Segments cultivated in distilled water supplied with 5 mmol dm−3 H2O2 for 8h (dark, constant 60 rpm shaking)	Sodium silicate 1 mmol dm−3
Ethanolb		Sodium silicate 1 mmol dm−3, 0.01% ethanol (v/v)

^a Final concentration of the substance in distilled water.

^b Treatments analysed by Raman spectroscopy and SEM-EDX for modelling the contributions of cell wall components to silica deposition.

^c After 4 h cultivation the segments were either collected immediately, or they were rinsed in distilled water for 10 min and cultured for additional 68 h in Si+ medium.

Raman microspectroscopy

Cross-sections were prepared from the root segments fixed in FAA either before the ex planta cultivation, after the cultivation, or at both time points. Several cross-sections from each root segment were placed on microscope slides, washed three times and mounted in distilled water, covered with coverslips, and sealed with nail polish to avoid water evaporation. Raman spectra were collected with an InVia spectrometer (Renishaw, UK) equipped with 532 nm laser, utilizing WIRE3.2 software (Renishaw, UK). Measurements were performed with 100 kW cm⁻² laser intensity, 0.1 s acquisition time and 150 accumulations per spectrum. At least five spectra from each root were collected, using at least three different roots per treatment. Collected spectra were smoothed (Savitzky–Golay algorithm, 9-point interval, polynomial order 4) and baseline corrected (adaptive baseline correction, coarseness 10%) using Spectragryph 1.0.7 (F. Menges ‘Spectragryph – optical spectroscopy software,’ Version 1.0.7, 2017, http://www.effemm2.de/spectragryph/).

Scanning electron microscopy and energy dispersive X-ray analysis

From each root segment fixed in FAA, we collected several hand cross-sections and a 0.5 cm-long piece, from which rhizodermis and outer cortical tissues were peeled-off (referred to as peeled-off segments; Lux et al., 2002). The samples were stuck onto a carbon tape and placed on a metal stand. Observations and energy dispersive X-ray spectroscopy (EDX) analyses were performed with a scanning electron microscope (JSM-IT 100 InTouchScope^TM; JEOL, Japan) under low vacuum (30 Pa), with accelerating voltage of 20 kV. Images were collected in back-scattered electron imaging mode. The EDX analyses were acquired from the peeled-off segments as point measurements with 20 s acquisition time per aggregate location, obtained from minimally 50 endodermal cells per treatment, from at least three different roots. Based on the EDX analyses the Si:C mass ratios were calculated.

Raman–EDX data analysis

Raman and EDX data collected in the ex planta experiments were used to estimate the contributions of cell wall components to silica deposition. Raman analysis is described in detail in Supplementary Dataset S1 at JXB online. Considering all the Raman spectra collected, peak positions (p) were identified (n_p=30; for assignments see Supplementary Dataset S1). Intensities of all the identified peaks were then extracted from the 72 h Si+ ex planta treatments to create a matrix of independent variables (the treatments used are indicated in Table 1, footnote b). For each variable, a mean value was calculated from the Si+ 1 mM treatment and used as a reference value. All observations in all treatments were afterwards expressed relative to corresponding reference values. For each treatment, a median of Si:C ratios (estimated based on EDX analyses) was then assigned as a dependent variable. The matrix of these relative peak intensities (independent variables) and relative Si:C ratios (dependent variable) was analysed using multiple linear regression with the weighted least squares estimation method. The number of independent variables was afterwards iteratively reduced using the backward elimination procedure, with maximal significance level for keeping a variable in the model being 0.005.

The analyses and their visual representations were performed using the Python programming language (version 3.5, Python Software Foundation, https://www.python.org/). The Si:C ratios were plotted using box plots showing interquartile ranges (IQR) with 95% confidence intervals indicated as notches and Tukey style whiskers (1.5×IQR) (Krzywinski and Altman, 2014). Identified outliers outside the 1.5×IQR whiskers are not displayed in the plotted data.

In vitro lignin synthesis by oxidative coupling and infrared spectroscopic analysis

An oxidative coupling reaction was carried out in several different substrate conditions: (i) CA alone, (ii) a 10/90 mixture of FA/CA, (iii) a 50/50 mixture of FA/CA, (iv) a 10/90 mixture of AX/CA, and (v) a 10/10/80 mixture of AX/FA/CA. For a reaction, 3 mg of substrate (equivalent to 4 mM of CA concentration in the reaction medium) was dissolved in 100 μl acetone and then added to phosphate buffer (3.7 ml, 0.1 M, pH 7.4), containing horseradish peroxidase (20 units, 20 μl) and hydrogen peroxide (30% solution, 4 μl). The solution was stirred at room temperature in the dark. Time points for ending the reaction were 3 and 18 h. The resulting suspension was rinsed with distilled water and centrifuged. The insoluble polymer was finally suspended in distilled water and dried, turning into an amorphous powder. Reactions were repeated with the addition of silicic acid (20 mM) for conditions in which a polymer was obtained. The reaction using CA only was also repeated with silicic acid concentrations of 10, 15, 20, 35, 50, and 80 mM. Silicic acid was obtained by adding 150 μl of tetramethyl orthosilicate to 850 μl of HCl 1 mM and stirred for 5 min.

Samples were weighed and prepared for infrared spectroscopic analysis by crushing the products to a fine powder in a mortar, mixing thoroughly with powdered KBr (about 1 part sample to 100 parts KBr by weight), and pressed to obtain a transparent disk. Transmission spectra were measured with a Nicolet 6700 FTIR spectrometer, in the range 400–4000 cm⁻¹, with 4 cm⁻¹ spectral resolution, collecting 40 scans. Spectra were visualized and analysed using OriginPro 2018 (64-bit) SR1 (OriginLab Corp., Northampton, MA, USA).

Results

Silica aggregates are formed along with cell wall deposition

In order to understand the relationship between silica supply and deposition, Si was provided in limited time intervals during seedling development (Fig. 1). Scanning electron micrographs collected at the back-scattered electrons mode could not detect silica deposits in plants grown without Si supply (hydroponics Si−). In plants supplied with Si throughout the entire cultivation (hydroponics Si+), massive silica aggregates formed, traversing most of the ITCWs, with tips protruding into the cell lumen. If Si was supplied only during the initial 3 d of cultivation (hydroponics Si+/Si−), the aggregates were substantially smaller, sunk within the ITCWs and covered with non-silicified cell wall layers. In plants supplied with Si only for the last 3 d of cultivation (hydroponics Si−/Si+), silica deposition occurred only within the inner (younger) portions of the ITCWs. The aggregates protruded to the cell lumen, similarly to aggregates in the hydroponics Si+ treatment, but with less cell wall material covering their surface. To quantify silicification we calculated the Si:C ratios of the aggregates, based on EDX analyses. In accordance with the scanning electron microscopy (SEM) images, the highest values of Si:C ratios were detected in the Si−/Si+ treatment, followed by Si+ and Si+/Si− treatment. No Si signal was detected in the Si− treatment (Fig. 1E). In the following experiments, we used the Si:C ratios accordingly, to estimate the silicification levels of silica aggregates.

Fig. 1.

Patterns of silica deposition in hydroponically grown seedlings supplied with Si in controlled time intervals. (A–D) Scanning electron micrographs of the endodermis in primary root cross-sections: (A) Si− treatment; (B) Si+ treatment; (C) Si+/Si− treatment; (D) Si−/Si+ treatment. E, endodermis; ITCW, inner tangential cell wall of endodermis; P, pericycle; C, cortical cell; arrowheads, silica aggregates. (E) SEM-EDX silicon to carbon ratios of silica aggregates exposed by removing the cortex tissue. Measurements were recorded from the radial direction, using peeled-off segments. Different letters above the boxes indicate statistically significant differences at P≤0.05.

The formation of silica aggregates requires metabolic activity

Based on Soukup et al. (2017), we established an ex planta silicification system as follows. Segments of seedling roots that were grown hydroponically without silicon for 3 d were cultivated ex planta for another 3 d in a 1 mM silicic acid solution. Bright regions in SEM images representing high silicon content were detected already 4 h after exposure to Si+ solution (Fig. 2). The deposition of the mineral was detected only in the inner, younger parts of the wall, similarly to intact roots developed in planta under Si−/Si+ treatment (Fig. 1D). After 3 d of cultivation, the Si:C ratios were also similar to aggregates developed in planta under Si−/Si+ growth, reaching 0.47±0.01 (Fig. 1E). This result shows that silica deposition is independent of the plant transpiration stream.

Fig. 2.

The time dynamics of endodermal silicification in detached root segments cultivated ex planta. Scanning electron micrographs of root segment cross-sections showing the endodermis before (A) and after (B–E) ex planta cultivation in Si+ 1 mM for (B) 4 h, (C) 24 h, (D) 48 h, and (E) 72 h. E, endodermis; ITCW, inner tangential cell wall of endodermis; P, pericycle; arrowheads, silica aggregates.

We further used ex planta cultivation to test whether silica aggregates form spontaneously as silicic acid reacts with cell wall materials. Root segments were collected from Si− plants, and their outer cortical tissues were removed mechanically (including the endodermal outer tangential wall). No silica deposition was detected after 72 h cultivation in Si+ 1 mM solution. Similarly, we found no mineral deposition in segments frozen in liquid nitrogen prior to ex planta cultivation. We thus concluded that silica is not forming as a simple reaction with cell wall constituents.

Next, we tested whether silica deposition involves metabolic activity of the root. We added to the ex planta growth medium 2,4-dinitrophenol (DNP), which is an ATP synthesis inhibitor. During a 4-h ex planta cultivation in 0.01 mM DNP, the aggregate formation was inhibited substantially. No silica deposition was detected under 0.05 mM DNP (Fig. 3). The segments were afterwards transferred to fresh Si+ medium without DNP and cultivated for an additional 68 h to reach the 72-h cultivation time in total. The silica deposition recovered after the transfer, suggesting that the components involved in silica deposition were not damaged by the DNP treatments. Decreasing the cultivation temperature to 4 °C hindered the silica deposition within a 4-h treatment as well (Fig. 3). Similarly to DNP treatment, transferring the segments back to regular cultivation temperature recovered the aggregate formation. Our results show that silica aggregates form only in live and active root tissue.

Fig. 3.

The effects of metabolic inhibitors on the aggregate Si:C ratio in ex planta cultivation. The treatment (DNP or low temperature) lasted 4 h (white symbols), after which the samples were moved to Si+ medium at normal cultivation conditions for additional 68 h (gray symbols). Different letters above the columns indicate statistically significant differences at P≤0.05.

Aiming to supply the segments with available energy and/or polysaccharide monomers we added to the growth solution sucrose, arabinose, or ATP (Fig. 4). As controls we used growth media containing only Si, at a concentration of 1 or 2 mM. Addition of sucrose resulted in faster silica deposition, as indicated by higher Si:C ratio along the aggregate formation (Fig. 4G). Moreover, the aggregates Si:C ratios were significantly higher in segments supplemented with sucrose as compared with segments cultivated without sucrose (Fig. 4H). In contrast, neither exogenous ATP nor arabinose increased the Si:C ratios (Fig. 4H). Our results thus propose that the sucrose provides glucose for the synthesis of the cell wall, which is a bottleneck for silica deposition.

Fig. 4.

Endodermal silicification in root segments cultivated in 1 mM Si (control) solution, supplemented with additional silicon (final concentration of 2 mM), sucrose, arabinose, or ATP. Scanning electron micrograph were collected after 72 h in vitro cultivation in control medium (A, B) and Si+ 1 mM and 1% sucrose (suc) (C, D) or Si+ 2 mM and 1% sucrose (suc Si 2mM) (E, F) medium. Panels (A, C, E) present cross-sections and panels (B, D, F) present longitudinal views (cortical tissues removed after cultivation). E, endodermis; ITCW, inner tangential cell wall of endodermis; P, pericycle; arrowheads silica aggregates. (G) The dynamics of silicification in detached root segments cultivated ex planta for 72 h, expressed as EDX Si:C mass ratios that were collected from the silica aggregates in longitudinal views, as shown in the panels (B, D, F). The shaded area represents 95% confidence intervals indicating statistically significant differences between the treatments, and the bars represent SD. (H) EDX Si:C mass ratios of silica aggregates after 72 h ex planta cultivation in control medium (1 mM Si), or medium containing 2 mM Si and 1% sucrose (suc Si 2mM), 1 mM Si and 1% sucrose (suc), 1 mM Si and 1% arabinose (arab), or 1 mM Si and ATP (1 or 5 mM). Different letters above the boxes indicate statistically significant differences at P≤0.05.

Silica deposition occurs together with the deposition of phenolic compounds

Raman microspectroscopy was used to characterize the endodermal ITWC. We found that the concentration of cell wall matrix polymers (lignin and hemicelluloses) were reduced significantly under ex planta as comparison to in planta growth. Nonetheless, Si supplementation did not change the mean Raman signal significantly. (Supplementary Dataset S1).

To test whether silica deposition is linked to specific components in the ITCW, we aimed to reduce specifically hemicelluloses or lignin, by treating root segments with a variety of drugs and chemicals (brefeldin A, ethanol, FA, salicylhydroxamic acid (SHAM), KI, ascorbic acid, and H₂O₂; Supplementary Dataset S1). A multilinear regression analysis was used to estimate the correlation between sets of ITCW Raman bands representing individual cell wall components and EDX Si:C ratios. An iterative backward elimination procedure (P_max=0.005) allowed us to reduce the model to the major components affecting silica EDX signals. The final model (R²=0.878, Prob(F)<0.001) identified 10 peaks that change significantly with the aggregate Si:C ratio (Fig. 5). A positive correlation was associated with vibrational bands of G-lignin and CA (1173, 1268, 1575, 1656 cm⁻¹), and negative correlations were associated with those attributed to H-lignin (643, 1205 cm⁻¹), S-lignin (1140, 1338 cm⁻¹), and phenolic aldehydes (1618 cm⁻¹). A positive effect was attributed also to the peak at 492 cm⁻¹, typically associated with a hemicellulosic C–O–C vibration. However, this contribution cannot be unequivocally assigned, since the characteristic frequencies of Si–O–Si vibrations are also located in this region.

Fig. 5.

Assessment of the contribution of cell wall components to the aggregate Si:C ratio based on Raman–EDX correlation analysis. A visual representation of multilinear regression coefficients after the backward elimination procedure with P_max=0.005. The combination of Raman peaks showing positive correlation with Si:C ratio indicates a dominant positive contribution of coniferyl alcohol (G-lignin constituent). The peaks exhibiting negative correlation are ascribed to phenolic aldehydes and H- and S-lignin.

Coniferyl alcohol polymers establish nucleation sites for silica deposition in vitro

To assess the potential relationship between silica and lignin deposition, we studied an in vitro lignin-like polymer production. Based on the Raman results we tested the hydrogenation of CA alone, or in mixtures together with FA and AX, using peroxidase and H₂O₂. No solid product was collected in the absence of CA, peroxidase, or H₂O₂. When the reaction was performed using CA alone, a dehydrogenated polymer was obtained with a reaction yield of 68% (Table 2). The infrared (IR) spectrum of the CA polymerization product (with bands at 1031, 1139, 1214, 1267, 1350, and 1598 cm⁻¹, band assignment in Table 3) is characteristic of lignin (Agarwal and Atalla, 2016). The presence of AX in the reaction strongly reduced the yield and did not allow for infrared analysis. Surprisingly, a 1:1 mixture of FA/CA also failed to produce any polymer (Table 2). However, conducting the reaction with a CA/FA ratio of 9:1 produced a lignin polymer modified by FA (Fig. 6A). The signal at 1750 cm⁻¹, assigned to the ester bond carbonyl stretching vibration, appeared as a consequence of the incorporation of FA into the polymer. Other spectral features, attributed to aromatic ring stretching, C–H stretching, and deformation vibrations were also found, evidenced by the bands at 1598, 1346, 1214, and 1139 cm⁻¹ (see Table 3 for assignments). When silicic acid was added at supersaturation (20 mM) to the oxidative coupling reaction of either pure CA or CA+FA, we identified infrared absorption bands typical of silica (Fig. 6B; Table 3). The chosen above-saturation concentration probably reflects the Si concentration at its deposition sites in the plant (Casey et al., 2004). During the reaction di- and tri-silicic acid as well as small oligomeric chains of silica may form in addition to lignin and silica particles. However, these small chemical species stay in solution, and cannot be recovered with the reaction product by centrifugation, in contrast to the lignin and bigger silica molecules.

Table 2.

Percentage yield of the in vitro lignin precipitation after 18 h

Substrate used	Yield (%)
CA	68
CA/FA (9:1)	65
CA+Si (20 mmol)	129
CA/AX (9:1)	22
CA/FA (9:1)+Si (20 mmol)	106
CA/FA/AX (8:1:1)	30
CA/FA/AX (8:1:1)+Si (20 mmol)	17

Yield of 100% equals the theoretical weight of polymerized phenolic (coniferyl alcohol, CA; ferulic acid, FA) and arabinoxylan (AX) precursors.

Table 3.

Assignment of the infrared bands in spectra obtained from the lignin-silica in vitro polymerization reactions

Peak position (cm−1)	Assignment
470	Si–O rocking mode
803	Si–O–Si symmetrical stretching mode
817	C–H out-of-plane in positions 2, 5, and 6, aromatic
854	C–H out of-plane in position 2, 5, and 6, aromatic
921	C–H out-of-plane, aromatic
966	–HC=CH out-of-plane deformation.
968	Si–OH stretching
1031	Aromatic C–H in-plane deformation; plus C–O deformation, in primary alcohols; plus C=O stretch (unconjugated)
1087	C–O deformation in secondary alcohols and aliphatic ethers
1093	Si–O asymmetrical stretching mode
1139	Aromatic C–H in-plane deformation; plus secondary alcohols plus C–O stretch
1214	C–C plus C–O plus C=O stretch
1267	C=O stretch
1330	Coniferyl ring
1367	Aliphatic C–H stretch in CH3, not in OMe; phenolic OH
1421	Aromatic skeletal vibrations plus C–H in-plane deformation
1463	C–H deformations in –CH3 and –CH2–
1506	Aromatic skeletal vibrations
1598	Aromatic skeletal vibrations plus C=O stretch
1650	H2O bending
1662	C=O stretch
2840–3000
2840
2871	C–H stretch in methyl and methylene groups
2935
3000
3399	O–H stretch

Assignment of vibrations is based on Yoshino et al. (1990), Faix (1991), Gunde (2000), Landreau et al. (2012), and Agarwal and Atalla (2016).

Fig. 6.

Infrared spectra of synthetic lignin, silica gel, and silica formed with the synthetic lignin. (A) Comparison of spectra of synthetic lignin formed using CA alone (lower line) and CA+FA (upper line) as precursors in the polymerization reaction. The main difference in the spectra is the presence of a band at 1750 cm⁻¹ that corresponds to a stretching vibration of the carbonyl (C=O) bond. This distinctive band indicates the presence of an ester bond and confirms the incorporation of FA into the polymer. (B) Infrared spectra of silica formed with the synthetic lignin produced using CA alone (lower line) and CA+FA (upper line) show typical silica bands. No differences were found between the spectra. (This figure is available in color at JXB online.)

Addition of silicic acid to the reaction solution resulted in an increased yield (Fig. 7). The increased yield relative to the CA precursors started only above 10 mM silicic acid, growing linearly with increasing concentration of silicic acid (Fig. 7A). The higher the silicic acid concentrations in the reaction, the higher was the fraction that precipitated (Fig. 7B). These results indicate a catalytic deposition, enhanced by silicic acid availability, as opposed to a stoichiometric reaction with the polymerized lignin. In agreement, when the silicic acid concentration was lower than 15 mM, infrared spectra of the product showed only typical lignin bands (Fig. 7C). When using concentrations of 15 and 20 mM silicic acid, we identified both silica and lignin bands, in agreement with minute silica yields (assuming that the yield of the CA polymerized was similar in these reactions). When using higher silicic acid concentrations, silica bands dominated the spectra and masked the lignin polymer signals (Fig. 7C).

Fig. 7.

Reaction yields of the in vitro lignin and silica formation. (A) Total yield calculated relative to the CA precursors added to the reaction. (B) Silica yield calculated by estimating a yield of the synthetic lignin of 68% (mean yield reaction with no addition of silicic acid). (C) Infrared spectra of the reaction product. In the products formed with concentrations of silicic acid higher than 20 mM, only the silica signals are visible and mask the lignin bands completely. (This figure is available in color at JXB online.)

Silicic acid autopolymerized from supersaturated solutions in reaction mixtures missing the lignin polymerization precursors (product denoted as SiO₂-Auto). This can be expected from the well-known ability of silicic acid to autopolymerize from solutions at concentrations above 2 mM (Iler, 1979). Nevertheless, the silica obtained together with CA polymer (denoted as SiO₂-CAp) exhibited substantial spectral and morphological differences from SiO₂-Auto (Fig. 8). Most prominently, a band that is assigned to the stretching vibration of non-bridging –SiOH (Wood et al., 1983) appeared in the spectra of SiO₂-CAp at 968 cm⁻¹ and in the SiO₂-Auto spectra at 983 cm⁻¹. SiO₂-Auto formed a soft transparent gel, similar to gelatine, which was still very soft after 4 h of drying at 40 °C. The gel hardened only after several days, preserving its transparent appearance. In contrast, SiO₂-CAp had whitish appearance with sandy texture, developing a solid consistency already after 4 h of drying (Fig. 8 inset).

Fig. 8.

Comparison of silica obtained by silicic acid autopolymerization (SiO₂-Auto) and silica precipitated together with the lignin like polymer (SiO₂-CAp). Infrared spectra of SiO₂-Auto (lower line) and SiO₂-CAp (upper line) show differences in the position of the band corresponding to the –SiOH stretching (shifted from 968 cm⁻¹ in SiO₂-CAp to 983 cm⁻¹ in SiO₂-Auto), the shoulder of the band at 1093 cm⁻¹ (asymmetric stretching Si–O–Si) and the relative intensity of the band at 803 cm⁻¹ (Si–O–Si symmetrical stretching mode). Such spectral variations indicate differences in the molecular order between the two materials. The SiO₂-Auto was formed at 20 mM concentration. The SiO₂-Cap was formed at 80 mM silicic acid, in order to minimize signals of lignin. Inset: photographs of the SiO₂-Auto (right) and SiO₂-CAp (left) suggest that these are two different minerals. Scale bars, 500 µm. (This figure is available in color at JXB online.)

Following the reaction in time we noted that while the lignin-like polymer started forming immediately, silica formation was delayed. We thus aimed to test whether the CA polymerization or the formed CA polymer caused silica deposition. After 3 h of CA polymerization in the presence of silicic acid, we could detect only the synthetic lignin and no silica (Fig. 9). SiO₂-CAp formed when the synthetic lignin (product of the 3-h experiment) was centrifuged, rinsed, redissolved in phosphate buffer with 20 mM silicic acid, and stirred for 15 additional hours. This was shown by the presence of prominent silica bands in the IR spectrum (Fig. 9, iii), consistent with the SiO₂-CAp IR spectra previously obtained (Figs 7C, 8). Our result indicates that the SiO₂-CAp formed at a later stage, catalysed by the CA polymer.

Fig. 9.

Infrared spectra of synthetic lignin and SiO₂-CAp at 3 and 18 h. (i) CA polymerization product in absence of silicic acid after reaction of 18 h. Synthetic lignin was detected. (ii) CA polymerization product in the presence of silicic acid after reaction of 3 h. Synthetic lignin with no silica bands was observed. (iii) After a 3-h reaction CA polymerization product was redissolved in a 20 mM silicic acid solution and left to react for another 15 h. Silica and lignin bands are clearly visible. (This figure is available in color at JXB online.)

Discussion

Silicification in grasses typically exhibits three general patterns: silica is deposited at epidermal surfaces, cell walls of internal tissues, or intercellular spaces, including the cell lumen voids (Kumar et al., 2017b). Whereas the first type of silicification may be driven by water evaporation and considered as passive/spontaneous, evidence indicate that at least some silica deposition is controlled actively (Perry et al., 1987; Kumar and Elbaum, 2017; Kumar et al., 2017a). In our previous work we found a positive correlation between the diameter of the silica aggregates in sorghum root endodermis and the extent of the thickening of the endodermal cell wall (Fig. 4S in Soukup et al., 2017), suggesting that the aggregates form only during tertiary wall deposition. Here we show that the formation of silica aggregates takes place exclusively in the cell walls of living cells (Fig. 1). The deposition only in new cell walls and the enhanced silicification with supplementation of sucrose support a model in which silica is deposited only as an integral part of the wall, during its formation (Fig. 4). The dependence of the mineral deposition on metabolic activity at the wall, and specifically deposition of matrix polymers, can explain the restricted final dimensions of silica aggregates (Sangster and Parry, 1976a,b; Soukup et al., 2014). We adapted SEM–EDX combined with Raman microspectroscopy to correlate the extent of silica and lignin deposition. SEM–EDX is a very common method to map silica (e.g. Sangster and Parry, 1976b), and relatively quantify it in a plant tissue (e.g. Lux et al., 2002). The Si/C ratio allowed us to assess surface silica deposition in relation to the background carbon deposition that exists in the cell wall. Raman mapping of plant tissues and specifically cell wall is a commonly used methodology (Gierlinger et al., 2012), which was applied to silicified plant tissues (Sapei et al., 2007; Gierlinger et al., 2008; Blecher et al., 2012). The statistical analysis of the spectral dataset opened this field to a broad range of applications (Schulte et al., 2008; Chylińska et al., 2014; Felten et al., 2015). Utilizing statistical tools, we were able to relate increased CA peaks to increased EDX-detected silica (Fig. 5).

Our in vitro experiments were conducted in supersaturated silicic acid solution. Above-saturation conditions are the standard conditions for studying silica bioproduction (Kröger et al., 1999; Dove et al., 2019). Supersaturation occurs in the xylem sap of members of the Poacae (grass family), including wheat (up to 8 mM; Casey et al., 2004), rice (up to 25 mM; Mitani and Ma, 2005), and sorghum (7–12 mM; our unpublished data). The formation of silica in the presence of polymerized CA suggests that the aromatic polymer stabilizes negative charges on oligomeric silicic acid (Dove et al., 2019). This is possible through hydrogen bonds forming between the aromatic hydroxyls and Si–OH groups. Bonds that remained active after the radical coupling may also stabilize negatively charged silicic acid. In contrast, the silica was not forming on CA monomers, nor during the CA polymerization, but only after most of the CA polymerized (Fig. 9). This result well conforms with the study by Fang et al. (2003) and Fang and Ma (2006), who demonstrate silica precipitation in vitro by synthetic or natural lignin, but not by lignin monomers, and raises the hypothesis that 3-D scaffolding is required for the mineral to grow.

Our previous analyses show that the sites of silica aggregation are low in lignin and rich in FA in comparison with the cell wall surrounding them (Soukup et al., 2017). We therefore tested in vitro the possibility of precipitating silica onto a CA–FA polymer. Infrared spectra suggested that the FA units bind to the polymer via ester bonds, indicative by the appearance of a peak at 1750 cm⁻¹ (Fig. 6). In contrast to the in vitro system, native silica aggregates autofluoresce blue under high pH (Soukup et al., 2014). This suggests that the FA in the silica aggregates is tethered to the wall via ether bonds (Leplé et al., 2007) and not through an ester. Silica formation may be induced similarly by the polymerization of monolignols in the apoplast of live cell walls. Root silica aggregates contain traces of FA and AX (Soukup et al., 2017) that may interact with the lignin to produce a silica deposition scaffold within the cell wall. We suggest a model in which negatively charged silicic acid is stabilized by newly synthesized cell wall lignin, accelerating silica nucleation and growth (Dove et al., 2019). Such interactions could be facilitated by hydrogen bonding between the hydroxyl groups on the surface of the lignin (on residues of CA, FA, other aromatic constituents, and possibly AX crosslinked to the lignin) and silica colloids or poly-silicic acid units. The stabilization may also involve bound H₂O molecules. Further research is needed to identify the moieties that are trapped in the silica and can nucleate its deposition.

From an evolutionary perspective, the role of silicification is still puzzling. Although many potential benefits were attributed either to the silicified cell walls (Hattori et al., 2003; Gong et al., 2006; He et al., 2014, He et al., 2015; Ma et al., 2015) or to the silica phytoliths (Massey and Hartley, 2009; Yamanaka et al., 2009; Hartley et al., 2015; Sato et al., 2016), a clear link to evolutional demands is missing (Strömberg et al., 2016). The interaction between silica and lignin we describe in this study recalls an important question – has the site-specific silicification evolved as a response to some environmental factors or as a protection against potential Si toxicity imposed by silica–lignin competition/co-precipitation? The question remains enigmatic and requires further studies.

Supplementary data

Supplementary data are available at JXB online.

Dataset S1. Supplementary protocols.

Glossary

Abbreviations:

AX

arabinoxylan

CA

coniferyl alcohol

DNP

2,4-dinitrophenol

EDX

energy dispersive X-ray spectroscopy

FA

ferulic acid

ITCW

inner tangential cell wall

SHAM

salicylhydroxamic acid