|Home | About | Journals | Submit | Contact Us | Français|
The herbivore defence system of true grasses (Poaceae) is predominantly based on silicon that is taken up from the soil and deposited in the leaves in the form of abrasive phytoliths. Silicon uptake mechanisms can be both passive and active, with the latter suggesting that there is an energetic cost to silicon uptake. This study assessed the effects of plant-available soil silicon and herbivory on the competitive interactions between the grasses Poa annua, a species that has previously been reported to accumulate only small amounts of silicon, and Lolium perenne, a high silicon accumulator.
Plants were grown in mono- and mixed cultures under greenhouse conditions. Plant-available soil silicon levels were manipulated by adding silicon to the soil in the form of sodium silicate. Subsets of mixed culture pots were exposed to above-ground herbivory by desert locusts (Schistocerca gregaria).
In the absence of herbivory, silicon addition increased biomass of P. annua but decreased biomass of L. perenne. Silicon addition increased foliar silicon concentrations of both grass species >4-fold. Under low soil-silicon availability the herbivores removed more leaf biomass from L. perenne than from P. annua, whereas under high silicon availability the reverse was true. Consequently, herbivory shifted the competitive balance between the two grass species, with the outcome depending on the availability of soil silicon.
It is concluded that a complex interplay between herbivore abundance, growth–defence trade-offs and the availability of soil silicon in the grasses' local environment affects the outcome of inter-specific competition, and so has the potential to impact on plant community structure.
Plants face a fundamental physiological trade-off between the allocation of limited resources to either growth and reproduction, or defence (Bazzaz et al., 1987; Herms and Mattson, 1992). The optimum value of this trade-off depends on two primary ecological factors: stress, e.g. shortages of light, water or mineral nutrients, and disturbance, e.g. intensity of competition or herbivore pressure (Grime, 1977; Coley et al., 1985; Siemens et al., 2010). Plants exhibit a great diversity of defensive adaptations against herbivores (Strauss and Zangerl, 2002), including physical barriers (Lucas et al., 2000; Hanley et al., 2007), toxic secondary metabolites (Rosentahl and Berenbaum, 1991; Bennett and Wallsgrove, 1994) and heavy metals (Rascio and Navari-Izzo, 2011). Since the expression of defence traits is often costly and can result in reduced growth and reproduction (Koricheva, 2002; Orians et al., 2010; Sletvold et al., 2010), many plants have evolved inducible defences that are only expressed following herbivore attack (Karban and Myers, 1989; Massey et al., 2007b), in addition to constitutive defences that are expressed regardless of damage.
Grasses (Poaceae) have long been thought to be relatively undefended and rely mainly on tolerance, i.e. the capacity to re-grow biomass lost to herbivores as their primary way of dealing with herbivory (McNaughton, 1979; Strauss and Agrawal, 1999; Del-Val and Crawley, 2005). Nevertheless, a variety of herbivore defence mechanisms have been indentified in grasses, including the formation of symbiotic associations with alkaloid-producing fungal endophytes (Clay, 1990; Schardl et al., 2004) and the production of toxic secondary metabolites (Vicari and Bazely, 1993). However, members of the grass family usually contain much lower concentrations of secondary metabolites than dicotyledons (Elger et al., 2009), suggesting that grasses depend on other defences to fight off herbivores (Grime et al., 1996).
It is now well established that grasses employ physical, silicon-based defences to counter herbivores and pathogens (Massey et al., 2006; Massey and Hartley, 2009; Reynolds et al., 2009). Silicon is taken up by roots in the form of monosilicic acid [Si(OH)4] (Guevel et al., 2007; Mitani et al., 2009) and is then transported primarily to the shoot epidermis, where it polymerizes to biogenic silica [SiO2·nH2O], forming opaline phytoliths (Kaufman et al., 1985; Prychid et al., 2003; Piperno, 2006; Mercader et al., 2009). Once deposited in this form, silicon can provide a physical defence based on the mechanical properties of opaline silica. High silicon concentrations in leaves correlate with high leaf abrasiveness (Massey et al., 2006; Massey et al., 2007a) which can increase tooth or mouthpart wear in both vertebrate (Jernvall and Fortelius, 2002) and invertebrate herbivores (Massey and Hartley, 2009; but see Kvedaras et al., 2009). Furthermore, the ability of herbivores to digest plant material effectively is reduced by the presence of silicon (Massey and Hartley, 2006), possibly because mastication and nutrient extraction are impeded by deposition of silicon (Hunt et al., 2008). In addition to serving as a mechanical defence, silicon plays a role in the induction of chemical defences against both herbivores (Correra et al., 2005; Gomes et al., 2005) and fungal pathogens (Cherif et al., 1994; Carver et al., 1998). Consequently, herbivores suffer reduced growth rates when feeding on high-silicon diets compared with low-silicon diets (Massey et al., 2006) and, given a choice, avoid feeding on high-silicon plants (Gali-Muhtasib et al., 1992; Massey et al., 2006, 2007a).
Effective herbivore defences benefit plants, but herbivore attack varies both spatially and temporally (e.g. Huitu et al., 2003; Yamasaki and Kikuzawa, 2003) and in the absence of herbivores an investment in herbivore defences may impose unnecessary costs on a plant, placing it at a disadvantage against its competitors (Strauss et al., 2002; Hanley and Sykes, 2009). To our knowledge, the metabolic cost of silicon-based defences has never been tested in this ecological context. The large differences in leaf silicon concentrations among grass species (Massey et al., 2007a) imply that the costs and benefits of silicon defences, and their impact on the competitive ability, are likely to vary. The modes of silicon uptake also differ among species: uptake can be active, passive or rejective (i.e. excluding silicon) (Takahashi et al., 1990; Currie and Perry, 2007). In comparison to chemical and carbon-based defences, silicon acquisition is thought to be metabolically inexpensive (Raven, 1983), but the presence of ATP-fuelled silicon-transporter systems, as identified in plant species such as rice (Ma et al., 2006), implies that there is an energetic cost to active silicon uptake (Raven, 1983). Furthermore, leaf silicon concentrations within a grass species are not static but can increase when the plant is under herbivore attack (Massey et al., 2007b), suggesting that there is a fitness cost associated with this defence. The magnitude of this herbivore-induced increase varies with plant species and is likely to be related to variation in uptake mechanism among species: those that take up silicon from soil actively may show a more pronounced increase in leaf silicon concentrations after damage than species that take up silicon passively.
Here, we report a study in which we investigated the ecological cost of silicon uptake in two common grass species (Poales: Poaceae): Poa annua (annual meadow grass) and Lolium perenne (perennial ryegrass), in the presence and absence of above-ground herbivory by desert locusts (Schistocerca gregaria; Orthoptera: Acrididae). Both grass species have a cosmopolitan distribution, and are native and common throughout Europe and Asia. Poa annua is an annual pioneer species most characteristic of disturbed habitats (Ellenberg, 1974; Hutchinson and Seymour, 1982). Like many other early-successional plant species, P. annua colonizes bare ground quickly and flourishes in the absence of competition. In contrast, L. perenne is a relatively strong competitor that is typically found in old pasture communities that are moderately grazed (Beddows, 1967; Ellenberg, 1974). Despite these apparent differences in growth strategies and habitat requirements, the two species frequently co-occur in nutrient-rich grasslands, trampled plant communities (e.g. turfs), and on gritted paths (Haeupler and Muer, 2000). Schistocerca gregaria naturally occurs in arid and semi-arid areas of Africa and south-western Asia, where it preferably feeds on grasses (Uvarov, 1977). Despite not occurring in the same habitat as the plant species used in this study, his species was chose t as a model generalist herbivore because it is known to cause significant damage to grasses and because it is readily available and easily cultured. Previous feeding preference trials have shown that food preferences of S. gregaria are very similar to that of other grass feeding herbivores [e.g. Spodoptera larvae (Massey et al., 2007a); field voles (Massey et al., 2007b)].
In a previous study, P. annua and L. perenne were found to accumulate a similar biomass after a growth period of 3 months, yet silicon concentrations in their leaves differed markedly, with those of L. perenne being more than twice than those of P. annua (Massey et al., 2007a). Lolium perenne has previously been reported to take up silicon actively (Jarvis, 1987), a mechanism believed to result in high leaf silicon concentrations (Cooke and Leishman, 2011). In contrast, the low leaf silicon concentrations typically found in P. annua suggest that active silicon uptake is of little importance in this species, or may not exist.
We hypothesized that in the absence of herbivores, increased plant-available soil silicon concentrations would lead to higher uptake rates and higher costs (expressed in lower biomass) in L. perenne than in P. annua, because L. perenne has previously been reported to accumulate much higher silicon concentrations in its leaves than P. annua (Massey et al., 2007a). Consequently, we expected the relative competitive ability of L. perenne to decrease under high soil-silicon availability when grown under interspecific competition. However, we expected that higher foliar silicon concentrations in L. perenne would justify their cost in the presence of herbivores by deterring herbivore feeding, and hence the predicted decrease in its relative competitive ability would not occur. A summary of predictions and measured outcomes is given in Table 1.
Three types of model plant communities, i.e. Poa annua monocultures, Lolium perenne monocultures and mixed cultures consisting of P. annua and L. perenne, were treated with two soil-silicon regimes (silicon added or not). The mixed cultures were treated with two herbivory regimes (herbivores added or not) to produce a total of eight different treatments. A total of six individual plants were planted into each community, with mixed cultures consisting of three individuals of each species. Each treatment was replicated ten times giving a total of 80 pots.
Seeds of P. annua L. and L. perenne L. were obtained from a commercial seed supplier (Herbiseed, Wokingham, Berkshire, UK) and germinated in vermiculite on 26 July 2009. After 5 d the seedlings were transplanted into plastic pots (18 cm diameter × 18 cm height) filled with peat. Peat was chosen as the growth substrate due to its low silicon concentrations, which provided a good control for treatments where additional silicon is added (Nanayakkara et al., 2008). Within a pot, six seedlings were planted at equal distance from the edge of the pot. In mixed-species pots, P. annua seedlings alternated with L. perenne seedlings. The plants were grown under greenhouse conditions (15–25 °C, 16 : 8 light : dark) in a randomized design for 6 months, until harvest in 20 January 2010. For the first 4 weeks the plants from the high soil-silicon treatments received 100 mL of a 500 mg L−1 sodium silicate (Na2SiO3·9H2O; Fisher Scientific, product number S/6400/60) aqueous solution every 4 d. From week 5 onwards, the plants received 100 mL of 1000 mg L−1 of sodium silicate solution every 4 d (modified after Massey et al., 2006). Plants from the low soil-silicon treatment received the same amount of tap water. Plants from all treatments received 100 mL of half-strength Hoagland's nutrient solution 2 and 4 weeks after germination, and 100 mL of full-strength Hoagland's solution 6 and 12 weeks after germination. Throughout the experiment all plants received tap water as required.
On 9 December 2009 (137 d after germination), all plants were enclosed in transparent perforated plastic bags. Half of the mixed-species pots from the low soil silicon (n = 10) and from the high soil silicon (n = 10) treatments were exposed to herbivory for the duration of 1 month by adding five second-instar nymphs of the locust Schistocerca gregaria obtained from a local pet shop to each pot. Locust numbers in each pot were censused at 3-d intervals, and dead or missing locusts were replaced with similar-instars from a stock culture. Locust mortality over the course of the experiment was around 30 %, but did not differ among treatments (data not shown) and was mainly caused by locusts being trapped between the pot and the perforated plastic bag. By the end of the experiment, all locusts were 4th instars except for one locust which reached the 5th instar stage.
After 1 month of exposure to herbivory, all plants from all treatments were harvested and the fresh above-ground biomass per species per pot was pooled and recorded. The plants were then dried in a fan-assisted oven at 70 °C for 1 week and the dry above-ground biomass per species per pot was recorded. The mean biomass of individual plants per pot was calculated by dividing the biomass per species per pot by the number of individuals of each species in the pot (three in mixed- and six in monocultures). The interspecific competitive balance between the two grass species was measured as the L. perenne : P. annua biomass ratio (hereafter LP : PA biomass ratio). All biomass data are presented and discussed as the mean above-ground dry biomass of individual plants per treatment.
Soil pH measurements were taken from five replicates of each treatment, chosen at random. Soil samples were collected after harvest and dried in a fan-assisted oven at 70 °C for 5 d. Then, 4 g of dry soil was thoroughly mixed in 50 mL of distilled water and allowed to settle. Soil pH measurements were taken by suspending a glass H+-sensing electrode attached to electronic pH meter (Hanna Instruments Ltd, UK) into the solution.
Dried plant material from each species from each pot was ground using a ball mill (Pulverisette 23; Fritsch GmbH, Germany). The ground plant material was pressed into pellets using a hydraulic press at 11 bars. Foliar silicon concentrations were analysed using a Niton XL3t portable XRF analyser (Thermo Fisher Scientific, Inc.), calibrated against silicon-spiked synthetic methyl cellulose (Sigma-Aldrich, product no. 274429). All analyses were performed in a helium atmosphere to avoid signal loss by air absorption and the samples were exposed to X-rays for a measurement time of 40 s. Four plant certified reference materials of known silicon content were employed for quality control measures (NCS DC73349 ‘Bush Branches and Leaves’, NCS ZC73013 ‘Spinach’, NCS ZC73014 ‘Tea’ and NJV 94-4 ‘Energy Grass’; National Analysis Center for Iron and Steel, China). All silicon data are presented as % silicon.
Foliar nitrogen was analysed using flash combustion of dried leaf samples (approx. 1·6 mg) followed by gas chromatographic separation (Elemental Combustion System; Costech Instruments, Inc.) calibrated against a standard of composition C26H26N2O2S.
The mean biomasses of individual plants per pot (biomass per species per pot/number of individuals per species per pot) from the non-herbivory treatments were analysed using a general linear model (GLM) with plant species (P. annua or L. perenne), competition type (mono- or mixed culture) and soil-silicon addition (silicon added or not) as fixed factors. Leaf silicon, water and nitrogen concentrations of plants from the non-herbivory treatments were analysed with a GLM with plant species, competition type and soil-silicon addition as fixed factors. The effect of soil-silicon addition on the soil pH was analysed using a t-test (silicon added or not). The LP : PA biomass ratio for plants from mixed cultures (herbivory and non-herbivory treatments) was analysed using a GLM with soil-silicon addition and herbivory (present or absent) as fixed factors. To test for the effects of herbivory on leaf silicon concentrations in the mixed-species cultures, a GLM was used with plant species, soil-silicon addition and herbivory as fixed factors. All statistical analyses were performed using the software package MINITAB (release 14, Minitab, State College, USA) and all data are presented as mean values ±s.e. If not stated otherwise, all differences reported are significant at P < 0·05.
Irrespective of competition type and soil-silicon addition, the above-ground biomass of L. perenne plants was approx. 25 % higher than that of P. annua plants (Fig.(Fig.11 and Table 2). Biomass of P. annua decreased by approx. 27 %, whereas biomass of L. perenne increased by approx. 28 % when grown in mixed cultures relative to monocultures (Fig. 1 and Table 2). The addition of plant-available soil silicon increased biomass of P. annua by approx. 56 %, but decreased biomass of L. perenne by approx. 22 % (Fig. 1 and Table 2), irrespective of competition type. The LP : PA biomass ratio was always greater than 1, showing that L. perenne was the dominant species in mixed cultures regardless of treatment (Fig. 2). Lolium perenne's superiority tended to decrease when silicon was added to to the soil, but this effect was not significant (Fig. 2 and Table 2).
Regardless of soil-silicon addition and competition type, leaf silicon concentrations of P. annua were approx. 50 % higher than those of L. perenne (Fig. 3 and Table 2). Silicon addition to the soil increased leaf silicon concentrations by over 400 % in both species, irrespective of competition type (Fig. 3 and Table 2). In monocultures, leaf silicon concentrations of L. perenne were 5·6 % higher than those in mixed cultures. In contrast, silicon concentrations of P. annua were 9·8 % lower in mono- than in mixed cultures (Fig. 3 and Table 2). Overall, leaf water concentrations did not differ between the two grass species (P. annua, 65·91 ± 0·53 %; L. perenne, 70·44 ± 0·50 %; both n = 60; Table 2). However, leaf water concentrations of L. perenne increased by 3·4 % when grown in mono- relative to mixed cultures (LP monoculture, 69·12 ± 0·74 %, n = 20; LP mixed culture, 71·10 ± 0·63 %, n = 40), whereas leaf water concentrations of P. annua decreased by 10·9 % when grown in mono- relative to mixed cultures (PA monoculture, 64·12 ± 0·68 %, n = 20; PA mixed culture, 66·80 ± 0·68 %, n = 40). The addition of silicon decreased leaf water concentrations by 5·3 %, irrespective of plant species identity and competition type (silicon added, 66·16 ± 0·42 %; no silicon added, 70·19 ± 0·62 %; both n = 60). Leaf nitrogen concentrations differed between species (P. annua, 1·24 ± 0·02 %; L. perenne, 1·38 ± 0·03 %; both n = 60; Table 2) and increased by approx. 10 % when silicon was added, irrespectively of species. Sodium silicate is a salt of a strong base and a weak acid and its aqueous solution can be highly alkaline. Consequently, silicon addition raised the soil pH from 4·4 (±0·04) to 6·4 (±0·06) (t38 = –26·8, P < 0·001).
There was no overall effect of herbivory on the LP : PA ratio (Table 2), but there was a significant interaction between silicon addition and herbivory: under low soil-silicon availability herbivory reduced the LP : PA ratio by approx. 30 %, while under high soil-silicon availability herbivory increased the LP : PA ratio by approx. 75 % (Fig. 2). The decrease in the LP : PA ratio under low soil-silicon availability was due to the preference of S. gregaria for L. perenne over P. annua (Fig. 4). Under high soil-silicon availability, P. annua was the preferred species (Fig. 4). Herbivory increased leaf water concentrations of both grass species by 3·6 % under low, but not under high, soil-silicon availability (Table 2). Herbivory had no effect on leaf silicon concentrations of P. annua (no herbivory, 2·09 ± 0·34 %; herbivory, 1·97 ± 0·31 %; both n = 20; Table 2), but increased leaf silicon concentrations of L. perenne by one-third (no herbivory, 1·29 ± 0·20 %; herbivory, 1·71 ± 0·27 %; both n = 20; Table 2), irrespective of the addition of plant-available silicon to the soil.
Silicon availability and plant competition in the absence of herbivores.
In the absence of herbivores, an investment in herbivore defences can impose unnecessary costs on a plant and place it at an immediate disadvantage against its competitors (Strauss et al., 2002; Hanley and Sykes, 2009). We thus hypothesized that increasing plant-available soil-silicon concentrations would have a strong negative effect on the biomass of the high silicon accumulator L. perenne that is known to take up silicon actively (Jarvis, 1987), but no or only a weak negative effect on the biomass of the supposedly low silicon accumulator P. annua, assuming there is a cost to silicon uptake in the absence of herbivory. In line with this hypothesis, the above-ground biomass of L. perenne decreased under high soil-silicon concentrations whereas that of P. annua increased.
Given our previous knowledge of leaf silicon accumulation by these species (Massey et al., 2006, 2007a) and that L. perenne takes up silicon actively, we expected its silicon levels to be higher than those of P. annua, but this was not the case: we found that regardless of soil-silicon concentrations, foliar silicon concentrations of P. annua were higher than those of L. perenne. Poa annua is known to exhibit large ecotypic variation (e.g. McElroy et al., 2004), which may explain the differences in foliar silicon concentrations between the present and previous studies. Indeed, in a recent experiment we found up to 2-fold differences in leaf silicon concentrations between genotypes of the same grass species, a phenonomena previously described for different ecotypes of rice (Deren et al., 1992). Further, transpiration rates are likely to have a large influence on silicon accumulation, due to the passive uptake of soil silicon employed by both active and passive silicon accumulators (Mitani et al., 2009). Interestingly, leaf silicon concentrations of L. perenne were higher in mono- than in mixed cultures whereas silicon concentrations of P. annua were lower in mono- than in mixed cultures. These results coincide with the effects of competition type on the leaf water content of the two species, suggesting that under the present conditions the water balance of the plants had a significant effect on silicon accumulation rates.
There are other ways silicon addition to the soil could have altered the relative performance of the two species since this study did not control for changes in soil pH and soil sodium concentrations caused by the application of sodium silicate. Since P. annua does not perform well in acidic soils (Ellenberg, 1974; Hutchinson and Seymour, 1982), the increase in soil pH caused by the addition of sodium silicate may benefit this species, but changes in soil pH cannot explain the differential effects on the two species, because L. perenne also grows optimally at neutral substrate pH (Beddows, 1967; Ellenberg, 1974). Also, both plant species had higher leaf nitrogen concentrations after silicon addition, suggesting that their uptake of nutrients was not adversely affected by soil conditions. Since both P. annua and L. perenne have previously been shown to respond in the same way to the application of sodium to the soil (Liem et al., 1985), changes in soil sodium concentrations caused by the addition of sodium silicate are also unlikely to have caused the observed differential effects on biomass of the two species.
The positive effect of plant-available soil silicon on the biomass of P. annua may be due to the fact that silicon plays an important role in the metabolism of many plant species (Isa et al., 2010), and although not strictly essential, it has been called a ‘quasi-essential’ plant nutrient (Epstein, 1999). For instance, silicon was found to enhance growth of cucumber plants and their resistance to fungal infection, notably by increasing the carboxylase activity of Rubisco (Adatia and Besford, 1986).
Regardless of the underlying mechanisms, our results indicate that plant-available soil-silicon concentrations can influence competitive interactions among plant species by promoting growth of some species, while reducing that of others. In natural systems, plant-available soil silicon is quickly taken up by plants and converted into phytoliths, before being returned to the soil in the form of plant litter (Farmer et al., 2005). On a landscape scale, concentrations of plant-available silicon are largely dependent on the weathering stage of the soil (Henriet et al., 2008), but to the best of our knowledge nothing is known about the spatial distribution of plant-available silicon in soils on a finer scale. Considering that soil pH has a large impact on phytolith dissolution (Farmer et al., 2005) and that plant species traits influence the uptake and restitution of soil silicon (Cornelis et al., 2010), plant-available silicon concentrations in soils are likely to be patchy. Herbivores often show large spatial and temporal variation in their abundance (e.g. Huitu et al., 2003; Yamasaki and Kikuzawa, 2003), and a complex interplay between plant-available soil-silicon concentrations, plant species traits (such as silicon uptake and restitution rates) and herbivore abundance may consequently impact on the composition of plant communities.
We hypothesized that in the presence of herbivores, the cost of active silicon uptake in L. perenne would be compensated for by the benefits received via increased herbivore resistance. Certainly L. perenne but not P. annua seemed to acquire increased resistance as soil-silicon availability increased: under low soil-silicon availability it was the preferred species but under high soil-silicon availability the locusts' feeding preference switched towards P. annua. This interactive effect between silicon availability and herbivory altered the interspecific competitive balance between P. annua and L. perenne, which concurs with previous results showing switches in host preference by a number of different herbivores under different soil-silicon regimes (Massey et al., 2006, 2007a, 2009). Even though none of the grasses flowered over the course of this experiment, silicon-induced switches in host plant preference may influence plant seed set, since herbivores such as field voles preferably attack the flowering culms of grasses (S. Reidinger, pers. obs.). However, it needs to be noted that the interplay between soil-silicon availability, plant growth and herbivore attack may be influenced by the developmental stage of the plant. Due to a larger root : shoot ratio, increased resource storages and a lower growth rate and metabolic activity, older dicotyledonous plants are usually better (chemically) defended than younger ones (Elger et al., 2009). Silicon concentrations in grasses increase with leaf age, but whether the trade-off between silicon-based defences and plant growth depend on the developmental stage of the plant and its species identity requires further investigation.
Both P. annua and L. perenne harbour fungal endophytes in their leaves (e.g. Prestidge and Gallagher, 1988; Bouton and Hopkins, 2003), and endophyte infection levels in L. perenne (Jensen and Roulund, 2004) and other grass species (Bazely et al., 1997) have been shown to correlate positively with grazing intensities. However, these correlations seem to be driven by a selective feeding behaviour of herbivores, which avoid feeding on endophyte-infected grasses (Bazely et al., 1997) and, consequently, favour the establishment and abundance of grass genotypes with high endophyte infection levels. Thus, differential effects of herbivory on endophyte infection levels of P. annua and L. perenne are unlikely to have caused the observed shift in herbivore feeding preference in the present study.
One unexpected result in the present study was that the high silicon concentrations in the leaves of L. perenne seemed to deter herbivore feeding, but the even higher silicon concentrations in the leaves of P. annua did not act as a deterrent, suggesting that silicon plays a more important role as an herbivore defence in L. perenne than it does in P. annua. The morphology of phytoliths and the type of tissue being silicified varies hugely among taxa (Piperno, 2006; Mercader et al., 2009). Very little is known about the functional significance of these differences (Cooke and Leishman, 2011), but they may contribute to interspecific differences in the effectiveness of silicon-based anti-herbivore defences among grasses. In contrast to perennial grass species (e.g. L. perenne), leaves of annual grass species (e.g. P. annua) often do not go through the full gradation of epidermal changes observed in grasses with a longer life cycle and the distribution of their phytoliths in the leaf blade is therefore less differentiated than in perennial species (Prat, 1951). Whether such differences can explain why silicon in L. perenne, but not in P. annua, seems to deter herbivore feeding requires further investigation.
Silicon is certainly not the only plant quality parameter affecting herbivore preference; indeed our previous work has found nitrogen concentration to be more significant in determining feeding behaviour of field voles (Massey et al., 2007a). Further, leaf water content has been shown to influence the palatability of plants to S. gregaria (Bernays and Lewis, 1986). Here we found leaf nitrogen concentrations to be higher in L. perenne than in P. annua, which may explain why under low silicon conditions the locusts preferably fed on L. perenne (Raubenheimer and Simpson, 2003). However, the observed effects of silicon addition on locust feeding preference can neither be explained by changes in leaf nitrogen nor water concentrations since these changed into the same direction and by the same magnitude in both grass species when silicon was added to the soil. The deterrent effect of a given level of silicon may also depend on other factors influencing the palatability of that species relative to another. For example, it may be that interspecies differences in other constituents (e.g. carbohydrate content, toughness, lignin or fibre) help explain why silicon appears to be a feeding deterrent in L. perenne but not P. annua. Though typically of low concentration in comparison to dicotyledonous species, grasses do contain secondary metabolites (e.g. phenolics) known to deter herbivore feeding (Grime et al., 1996; Elger et al., 2009). Thus, it is possible that L. perenne was less susceptible to locust herbivory than the more silicon-rich P. annua, because it was able to mobilize such chemical defences in response to herbivore attack. In light of the recent advances in metabolomic profiling techniques, future work is needed to decipher the role of chemical defences in grasses, particularly since qualitative differences in the composition of phenolics have recently been shown to be more important in respect to herbivore food choice than total phenolics concentrations (Czerniewicz et al., 2011).
Locust feeding increased leaf silicon concentrations of L. perenne by one-third, supporting previous findings that showed silicon to be an inducible defence in this plant species (Massey et al., 2007b). Changes in the plants' water balance following herbivory are unlikely to have caused these effects since there was no species-specific effect of herbivory on leaf water concentrations in the two species. Instead, we suggest that an herbivory-induced increase in the active soil silicon-uptake system of L. perenne, but not of P. annua which may lack such a system, led to the higher leaf silicon concentrations when the plants were under attack. The grass Molinia caerulea has been shown to respond to artificial damage of its leaf tips with an abnormal accumulation of phytoliths in leaf cell types which are not usually silicified (Parry and Smithson, 1964). It is thus possible that herbivore-induced increases in leaf silicon concentrations are accompanied by changes in the distribution of phytoliths within the leaf blade and/or changes in phytolith structure, which in turn may influence herbivore food choice.
In conclusion, the results of this study show that the availability of soil silicon mediates interspecific competition among plants via two different mechanisms: directly through its species-specific effects on plant biomass, and indirectly through altering herbivore attack rates. Differences in the ability of plant species to increase uptake of silicon, particularly in response to herbivore attack, may have consequences for their competitive ability. Hence, heterogeneous distributions of soil available silicon have the potential to impact on plant community structure, particularly in the presence of herbivores.
This work was supported by the Natural Environment Research Council to S.E.H (NE/F003137/1). We thank Rene Eschen and Jennifer Rowntree for comments on the manuscript.