Study species and phylogeny
We focused on 37 species from 24 genera, 18 families and 14 orders, including 13 congeneric and three confamiliar species pairs, and five additional species (Table S1). Species were selected based on the following criteria: (1) species pairs included contrasting moisture associations based on Ellenberg’s indicator values [moisture (M)-value; Ellenberg et al. 2001] with M-values differing by at least two levels within pairs, and M-values for species from drier and moister habitats ranging from 3 to 5 and 5 to 8, respectively; (2) maximizing phylogenetic diversity, i.e., choosing species from different taxonomic orders; (3) species associated to open habitats (i.e., we excluded species of shaded habitats); and (4) commercial availability of seeds (Rieger-Hofmann GmbH, Blaufelden, Germany). Seed production aimed to maintain the genetic diversity of the species' source populations in southern Germany.
We generated a phylogenetic tree of the study species based on the highly resolved species-level megaphylogeny by Zanne et al. (2014). For six study species, which were not included in Zanne et al. (2014), we used phylogenetic data of congeneric species as replacement (for details see Table S1). We generated the phylogeny using the R-package “ape” (Paradis and Schliep 2019).
Five individuals per species and soil Si treatment were included (see Table S1 for a few exceptions). Seeds were germinated in multicell plug trays with low Si soil in a greenhouse (Bayreuth, Germany). Seedlings were individually transplanted into separate pots (Deepot Cells, Stuewe & Sons, Oregon, USA; diameter: 6.5 cm, depth: 36 cm) with the experimental substrates (see below). Seedlings of all species were transplanted within 1 day at cotyledon stage or after development of first foliage leaves.
The experimental substrates consisted of 60% sieved sandy topsoil, 30% quartz sand and 10% clay granulate (hereafter referred to as soil). For high Si availability the soil was enriched with 12.5 g of amorphous Si (Aerosil 300, Evonik Industries AG, Essen, Germany) per litre substrate and thoroughly homogenized. Addition of Aerosil 300 does not lead to changes of soil pH (Schaller, unpublished data). In the low Si treatment, no Si was added. The resulting plant-available soil Si concentration was 110 mg kg−1 for the high Si soil and 40 mg kg−1 for the low Si soil (for analysis see below), equivalent to high and intermediate levels of plant-available Si in agricultural soils (Caubet et al. 2020).
We watered all plants regularly by hand and fertilized them four times during the experiment (0.3% NPK fluid fertilizer, Wuxal Super, Wilhelm Haug GmbH & Co. KG, Ammerbuch, Germany). Temperatures ranged between 18 and 22 °C and plants grew under natural light intensity supplemented with artificial greenhouse light (Plantstar 400 W E40, Osram, Munich, Germany). The position of species and treatments in the greenhouse was randomized and rearranged regularly.
We harvested the aboveground biomass of the plants 10 weeks after transplantation in a randomized order. Leaves of each individual were cleaned, removing any residual soil material, oven-dried for 48 h at 65 °C, and ground for analyses.
Foliar and soil Si analyses
Silicon was extracted from the leaves for 5 h by an alkaline method using 30 mg of leaf material and 30 ml of 0.1 M sodium carbonate solution (Na2CO3) in a regularly shaken water bath following Struyf et al. (2010). The solution was subsequently passed through a 0.2 µm syringe filter (ChromafilXtra CA-20/25). Soluble soil Si was extracted in CaCl2 following Schaller et al. (2018). Three g of soil were shaken with 30 ml of 0.01 M CaCl2 for 1 h at ambient laboratory temperature. The suspension was centrifuged (8000×g, for 10 min) and the supernatant decanted.
The Si concentration of the leaf or soil extract was determined with inductively coupled plasma optical-emission spectrometry (ICP-OES) using a Varian Vista-Pro Radial element analyser (Varian Inc., Palo Alto, USA), compare Schaller et al. (2018).
Interspecific differences in foliar Si, and intraspecific responses to soil Si
Interspecific differences of foliar Si across all 37 species grown on the high Si soil were analysed using one-way ANOVA with a Tukey HSD post-hoc test using the R-package “agricolae” (de Mendiburu 2020). Differences of foliar Si within each of the 16 congeneric or confamiliar species pairs were analysed using one-way ANOVA. To analyse the effect of soil Si availability on foliar Si concentrations we calculated a two-way ANOVA with species (26 species), treatment (low vs. high Si) and their interaction as explanatory factors. Furthermore, we tested for differences of foliar Si concentrations within each individual species by one-way ANOVA. Foliar Si concentrations were log10 transformed to fulfil normality and homoscedasticity. We excluded one outlier, an individual of the high Si treatment (Verbascum lychnitis), which we were able to trace back to soil contamination.
To evaluate whether intraspecific responses of foliar Si to soil Si are related to their foliar Si concentration under high soil Si availability, representing the species’ maximum Si accumulation capacity, we calculated a response ratio for each species based on (untransformed) mean values of foliar Si in each treatment (RRFoliar Si; compare Hedges et al. 1999) as: RRFoliar Si = log10 (foliar Sihigh Si/foliar Silow Si). A more positive RRFoliar Si indicates a higher increase of foliar Si concentrations in response to higher soil Si availability. We tested the relationship between the species’ RRFoliar Si and their foliar Si concentrations in the high Si treatment with Spearman rank correlation (n = 26). To assess whether species ranking of Si concentrations stays consistent under different soil Si availabilities, we additionally calculated the Spearman rank correlation coefficient (ρ) of foliar Si concentrations between the low and high Si treatment (n = 26).
To ease data interpretation, we classified species based on their foliar Si concentrations under high soil Si availability into low-accumulating species (< 5 mg g−1), assumed to predominantly take up Si by passive diffusion, and high-accumulating species (> 5 mg g−1), assumed to additionally take up Si actively through ATP-consuming Si transporters (compare Strömberg et al. 2016 for a similar approach; threshold values based on Ma et al. 2001).
Phylogenetic signal in foliar Si
To analyse the phylogenetic signal in the species’ mean foliar Si concentrations in the high Si treatment (i.e., their maximum accumulation capacity), and in RRFoliar Si, we separately calculated Pagel’s lambda λ (Pagel 1999) for each parameter (across 32 and 26 species, respectively) using the function “phylosig” from the R-package “Phytools” (Revell 2012). A Pagel’s lambda λ = 1 indicates that phylogeny can fully explain the variation in foliar Si concentrations (or RRFoliar Si), i.e., a pure Brownian model of evolution, and λ = 0 indicates that phylogenetic relationships cannot explain the variation. We separately tested the null hypotheses of λ = 0 and λ = 1 using likelihood-ratio tests.
Relationship of foliar Si to species moisture association
We tested the relationship between species’ association to habitat moisture (M-value, Ellenberg et al. 2001) with foliar Si (under high Si) across 32 species (all species pairs) and with RRFoliar Si across 26 species. We calculated ordinary least square regressions (OLS), and additionally conducted phylogenetic generalized least square regressions (PGLS) using the R-packages “nlme” (Pinheiro et al. 2020) and “ape” (Paradis and Schliep 2019) to account for species phylogenetic relatedness. We used two variance–covariance structures in the PGLS models, each representing a potential evolutionary trajectory along the branches of the phylogeny; either evolution by strict Brownian motion or Brownian evolution adjusted by Pagel’s λ as scaling parameter (Pagel 1999), with the scaling parameter λ obtained using maximum likelihood. Since the models based on strict Brownian motion consistently showed the poorest fit (highest AIC values), we only included the results of λ-modified PGLS models.
We additionally tested the relationship of foliar Si to the species’ association to habitat moisture separately for the subset of species pairs containing only low-accumulating species, and for the subset of species pairs containing at least one high-accumulating species, using OLS and PGLS regression.
All statistical analyses were performed in R version 4.0.2 (R Core Team 2020).