Tall fescue (Festuca arundinacea) is a common pasture grass and a high Si accumulator (Hodson et al., 2005). Seeds of tall fescue (accession T 9627), obtained from Margot Forde Germplasm Centre (Palmerston North, New Zealand), were sterilised in a solution of 0.9% sodium hypochlorite and 0.1% Triton X-100 for 30 min, followed by washing several times with water before being inserted into perlite irrigated with water. Seeds were stratified at 4°C for 3 days, and plants were grown in the glasshouse for 2 weeks using a rectangular plastic tray to achieve uniform seedling growth. Two weeks after germination, individual plants were transferred to hydroponics cups. The hydroponics setup consisted of two nested disposable plastic cups as per Hall et al. (2019). Each cup was filled with approximately 350 mL of full-strength standard hydroponic solutions following the protocol of Jung et al. (2015). Seedlings were grown in three plant growth chambers (TPG-1260TH, Thermoline Scientific, NSW, Australia), maintained at a reduced CO₂ of 200 ppm, an ambient CO₂ of 410 ppm, and an elevated CO₂ of 640 ppm CO₂ concentrations; the latter CO₂ concentration predicted for 2100 under the RCP6.0 scenario outlined by the IPCC (2014). Chambers were illuminated with five 400-W Sunmaster Dual Spectrum High-Pressure Sodium globes at 350 µmol m⁻² s⁻¹ at the plant canopy level. Daytime air temperature was regulated at 26°C and fell to 18°C at night on a 15L:9D photoperiod cycle. Humidity was controlled at 50% ( ± 6%). Carbon dioxide within the chambers was monitored by a Li-Cor LI-820 CO₂ gas analyser, with CO₂ (food grade, Air Liquide, NSW, Australia) injected from pressurised cylinders through solenoid valves. For reduced CO₂ treatment in the chamber, the computer controller constantly monitors CO₂ and powers fans to direct chamber air through the scrubbers filled with Sodasorb^® (W.R. Grace & Co, Chicago, USA).

The experimental design consisted of 252 hydroponically grown tall fescue plants. The experiment comprised a factorial combination of CO₂ concentrations (200, 410, or 640 ppm), Si (+Si or −Si), and herbivore (herbivory, +H; no herbivory, −H) treatments (see Figure 1 ). Si treatments (+Si) used liquid potassium silicate (K₂SiO₃) (Agsil32, PQ Australia, SA, Australia) at a concentration equivalent to 2 mM SiO₂. Chemically, silicic acid polymerises to form silica gel when the concentration of silicic acid exceeds 2 mM (Ma and Yamaji, 2006). Potassium chloride was added to the control (−Si) cups to balance additional K⁺ ions in +Si treatments. The pH of both +Si and −Si solutions was adjusted to 5.6 using hydrochloric acid to reduce silicate polymerisation (Ma and Yamaji, 2006). Solutions were replaced weekly for the first 2 weeks and then three times a week afterwards. Cups were rotated and chambers were swapped weekly to minimise chamber effects and pseudo-replication as previously described by Johnson et al. (2018). Plants were grown hydroponically for a further 6 weeks (42 days, Figure 1 ) before insect inoculation.

To assess the impacts of different CO₂ concentrations and Si supplementation on the growth of H. armigera larvae, a feeding performance assay was conducted. Helicoverpa armigera third instar larvae supplied by CSIRO Agriculture & Food, Narrabri, Australia, reared on an artificial diet (Teakle and Jensen, 1985), were used for feeding assays. Initially, larvae were starved for 24 h and weighed, and a single larva was either applied for each plant shoot [herbivory (+H)]) or kept control [no herbivory (−H)] (see Figure 1 for details). Each plant was then caged with transparent Perspex sheaths and herbivores were allowed to feed on shoot parts of the plants for 6 days, after which they were removed and starved for a further 24 h to allow the frass to pass, before being reweighed. All frass were collected, dried, and weighed. Frass production was used as a surrogate for plant consumption. RGR was calculated according to Massey and Hartley (2009). RGR estimates the change in larval fresh mass relative to initial mass and was calculated as mass gained (mg)/initial mass (mg)/time (days).

Immediately after herbivore removal, plants from all treatment groups were harvested into liquid nitrogen and stored at −80°C until analysis. For the measurement of enzymatic activities, ca. 0.05 g of leaf tissue was ground in liquid N and homogenised in 3 mL of ice-cold 100 mM K-phosphate buffer (pH 6.8) containing 0.1 mM EDTA. The homogenate was centrifuged at 16,000 g for 15 min and the supernatant was used as the source of crude extracts. The supernatant was utilised to measure the activity of antioxidant enzyme such as CAT, APX, and SOD.

CAT activity was measured spectrophotometrically following the method of Fimognari et al. (2020) and Maksimović and Živanović (2012). Reaction mixture consists of 50 mM potassium phosphate buffer, pH 7.0, 20 mM H₂O₂, and crude extract. Absorbance at 240 nm was recorded for 130s using CLARIOstar^® plate reader (BMG Labtech, Ortenberg, Germany) in 96-well plates (96-well Flat Bottom Plate, Greiner Bio-one, Australia). For control reactions, H₂O₂ was omitted. Enzyme activity was calculated using the molar extinction coefficient 36 × 10³ mM⁻¹ m⁻¹ and expressed as µmol H₂O₂ oxidised g⁻¹ FW min⁻¹.

APX activity was assayed according to Hartmann and Asch (2019). The reaction mixture consists of 50 mM potassium phosphate buffer, pH 7.0, 0.2 mM ascorbate, 0.2 mM H₂O₂ and crude extract. Absorbance at 290 nm was recorded for 130 s using CLARIOstar^® in 96-well plates. APX activity was calculated according to Hartmann and Asch (2019); one unit of APX is defined as the amount of enzyme required to oxidise 1 μmol of ascorbic acid per minute.

SOD activity was estimated following the inhibition of photochemical reduction of nitroblue tetrazolium (NBT) by the enzyme according to Hartmann and Asch (2019). The reaction mixture contained 0.05 M sodium carbonate, 13.3 mM methionine, 1.3 μM riboflavin, 21 μM NBT, and plant extract (Hartmann and Asch, 2019). The reaction took place in a chamber under illumination of a 30-W fluorescent lamp at 25°C. The reaction was started by turning the fluorescent lamp on and stopped 5 min later by turning it off. The blue formazan produced by NBT photoreduction was measured as increase in absorbance at 560 nm. The blank solution had the same complete reaction mixture but was kept in the dark. One SOD unit was defined as the amount of enzyme required to inhibit 50% of the NBT photoreduction in comparison with wells lacking the plant extract and expressed as units of enzyme activity g⁻¹ FW min⁻¹ (Cavalcanti et al., 2004).

Harvested sample leaves were oven-dried for 3 days at 60°C and ball milled for further analysis (see sample size from Figure 1 ). For foliar Si analysis, roughly 80 mg of ground leaf material was analysed using x-ray fluorescence spectrometry (Epsilon 3^x, PANalytical, EA Almelo, The Netherlands) as per Reidinger et al. (2012). Si measurements were calibrated against a certified plant reference material of known Si concentrations (Hiltpold et al., 2017). For foliar C and N concentration, approximately 7 mg of ground leaf material was analysed using Elementar Vario EL Cube, CHNOS elemental analyser (Elementar Analysensysteme GmbH, Hanau, Germany), at a combustion temperature of 950°C.

All data were analysed using SPSS (version 27) statistical software. Before analysis, all data were checked for assumptions of normality for residuals according to inspection of quantile–quantile plots. Plant dry mass was analysed on square-root transformed data whereas CAT, C:N ratio, and RGR were analysed on log10 transformed data, as they did not meet the assumptions of normality. Foliar Si was analysed using two-way analysis of variance (ANOVA) type = II, comparing CO₂, and herbivory (larval fed vs. undamaged controls) as treatments and their interaction. For foliar Si analyses, control (−Si) plants were omitted since −Si plants had Si concentrations lower than the machine detection limits (Reidinger et al., 2012). Plant dry mass, antioxidant enzyme activities (CAT, APX, and SOD), foliar C, N, and C-to-N ratio concentrations were all analysed using three-way ANOVA type = II, comparing CO₂, Si (Si-supplemented vs. non-Si-supplemented plants) and herbivory as treatments and their interaction. Additionally, we tested the independent effects of CO₂ on antioxidant enzyme using a one-way ANCOVA, with CO₂ levels as a fixed factor and foliar Si concentration fitted as a covariate. For herbivore RGR and frass produced, three insects escaped, so data were analysed using type = III ANOVAs due to the unbalanced design. Bonferroni post hoc test (Aslam and Albassam, 2020) was applied for pairwise multiple comparisons when interaction terms were statistically significant. Potential relationships between foliar Si and herbivore RGR, frass produced, CAT, APX, and SOD enzymes activity were investigated using Spearman’s rank correlation test.

Averaged across CO₂ treatments, Si supply increased plant dry mass by 160% relative to those grown without Si supply, whereas elevated CO₂ increased plant dry mass by twofold and threefold compared to reduced CO₂ and ambient CO₂, respectively ( Figure 2A ; Table 1 ). Furthermore, herbivory decreased plant dry mass in Si-free (control) plants by 1.5-, 2-, and 2-fold under reduced, ambient, and elevated CO₂, respectively, compared to Si-supplemented plants ( Figure 2A ). Si supplementation decreased foliar C concentrations under all CO₂ regimes. This effect was reversed when herbivores were present and foliar C concentrations increased to levels observed in Si-free plants ( Figure 2B ; Table 1 ). Si supply decreased foliar C by 149%, 177%, and 107% under reduced, ambient, and elevated CO₂, respectively, regardless of herbivore treatments ( Figure 2B ). Reduced CO₂ significantly decreased foliar C ( Figure 2B ; Table 1 ). Si supply decreased foliar N under all CO₂ regimes. However, there was also an effect of CO₂ whereby reduced CO₂ increased foliar N by twofold and threefold compared to ambient and elevated CO₂, respectively, irrespective of herbivore treatments ( Figure 2C ; Table 1 ). In addition to variations in C and N concentrations, there was also an effect on their ratio. Si supply increased foliar C:N ratio especially when plants were damaged by herbivores or in herbivore-free plants only under elevated CO₂ ( Figure 2D ; Table 1 ). While herbivory increased foliar C:N ratio regardless of CO₂ levels, reduced CO₂ decreased foliar C:N ratio relative to elevated CO₂ ( Figure 2D ; Table 1 ). Foliar Si accumulation [% dry weight (DW)] was significantly higher under reduced CO₂ relative to ambient and elevated CO₂ ( Figure 3 ; Table 1 ).

CAT activity increased in response to Si supply and herbivore damage overall, although this was only apparent under reduced CO₂ concentrations ( Table 1 ; Figure 4A ). The significant interaction between Si and CO₂ reflects that Si impacts were only apparent under reduced CO₂ concentrations ( Table 1 ). In contrast, Si supply increased APX enzymes activity under all CO₂ regimes and SOD enzyme activity under reduced and ambient CO₂ ( Table 1 ; Figures 4B, C ). Herbivory caused higher APX activity specifically in Si-supplemented plants ( Table 1 ; Figure 4B ); however, it had no significant effect on SOD enzyme activity ( Table 1 ; Figure 4C ). Overall, antioxidant enzyme activity (CAT, APX, and SOD) declined with increasing CO₂ concentrations ( Table 1 ; Figures 3A–C ). Including foliar Si as a covariate in ANCOVA indicated that the changes in antioxidant enzyme activity were linked to CO₂ levels, which fully explained the observed changes in CAT (F _{1,53 =} 11.67, p = 0.001) and APX (F _{1,42 =} 5.79, p = 0.021) but not in SOD (F _{1,53 =} 1.130, p = 0.258). There was a positive correlation between foliar Si concentrations and concentration of CAT under reduced CO₂, and concentration of APX under elevated CO₂ ( Figures 5A, B ). Interestingly, frass produced was negatively correlated with SOD (r = −0.310, p = 0.038), but had marginally non-significant negative correlation with APX (r = −0.286, p = 0.057). However, there was no such relationship observed between CAT and frass produced (r = −0.191, p = 0.208) (data not shown).

Si supplementation decreased RGR under all CO₂ regimes; RGR was significantly lower under reduced CO₂ compared to elevated CO₂ ( Table 1 ; Figure 6A ). Si supply decreased the amount of frass produced by caterpillars (indicative of feeding efficiency) under all CO₂ levels; CO₂ had no significant effect on frass production, although there was a large increase in production in Si-free plants grown under elevated CO₂ ( Table 1 ; Figure 6B ). While herbivore RGR and frass produced were negatively correlated with foliar Si under ambient CO₂, there was no such relationship observed under the other two CO₂ regimes ( Figures 6C, D ). Here, the correlation between rate of herbivore feeding on foliar tissues of tall fescue grown under different CO₂ concentrations potentially indicates a new perspective towards mitigating challenges of CO₂ enriched environment on plant defences.

The effects of CO₂ in this study on Si concentrations are broadly similar to the few studies exploring this, whereby elevated CO₂ leads to decreased Si accumulation (Ryalls et al., 2017; Johnson and Hartley, 2018; Biru et al., 2022; Johnson et al., 2022; Johnson et al., 2023), whereas reduced CO₂ leads to increased Si accumulation. To the best of our knowledge, the current study is only the third study to investigate the latter (Biru et al., 2020; Biru et al., 2021). Si has been hypothesised to act as a structural substitute for C at a lower metabolic cost, particularly when CO₂ concentrations were lower in the Miocene (Craine, 2009).

Exogenous application of Si has been shown to increase antioxidant enzyme including CAT, APX, and SOD (Kim et al., 2016; Hasanuzzaman et al., 2018; Ahanger et al., 2020). It seems likely that changes in Si in response to CO₂ growing conditions influenced activity of antioxidant enzyme in the current study. Additionally, the ANCOVA results indicated that CO₂ affects antioxidant enzyme activity (e.g., CAT and APX), which were mostly explained by the direct impacts of CO₂ levels on antioxidant enzyme. However, the observed positive correlation of leaf Si with CAT and APX under reduced CO₂ and elevated CO₂, respectively, suggests that CAT defence response may be linked to higher levels of Si under reduced CO₂, whereas APX defence response may be associated with increased induction of defence responses following herbivore attack under elevated CO₂. Overall, our results demonstrated significant augmentation of antioxidant enzyme responses via increased Si uptake under reduced CO₂ as well as by the direct effect of CO₂ levels (see Figure 7 ). In contrast, previous studies have reported that elevated CO₂ increases activity of antioxidant enzyme in different plants (Wang et al., 2003; Moghimifam et al., 2020). For example, Moghimifam et al. (2020) found that elevated CO₂ enhances CAT, polyphenol oxidase (PPO), SOD, and proline activity in algae (Dunaliella sp.). Our result may reflect that reduced CO₂ often increases photorespiration (Moroney et al., 2013; Voss et al., 2013) and since photorespiration is the key source for ROS production (Voss et al., 2013), reduced CO₂ may cause increased antioxidant enzyme activity in order to scavenge excessive ROS produced. However, these studies did not address whether Si played a role in these changes, while our findings suggest Si as an important influencer of the activity of these enzymes.

In addition to CO₂ having direct and Si-mediated impacts on antioxidant enzyme activity, it is also possible that the amount of herbivore damage played a role in antioxidant enzyme activity. Herbivores induced higher activity of CAT and APX, so these enzymes activity is at least associated with the levels of damage done to the plant, which has been similarly reported in previous studies (Leitner et al., 2005; Yang et al., 2017). Increased levels of antioxidant enzyme under reduced CO₂ may reflect the fact that insects were doing less damage to the plant under these conditions and, therefore, there was less ROS oxidative stress that may have persisted in plant tissues to react with higher levels of antioxidant enzyme. Results from the ANCOVA and correlation tests also revealed that CO₂ directly changed antioxidant enzyme activity and altered feeding efficiency (leaf consumption), which eventually reduced the growth rate of herbivores.

The recent finding by Biru et al. (2021) showed that H. armigera RGR was lowest when fed on the model grass Brachypodium distachyon grown under reduced CO₂ due to these plants having higher levels of Si defences compared to plants grown under ambient and elevated CO₂ concentrations. In the current study, we observed that tall fescue had the highest concentrations of foliar N when grown under reduced CO₂, which, in theory, could have promoted herbivore RGR because N is frequently the limiting factor in insect herbivore diets (Mattson, 1980; Huberty and Denno, 2006). The lower production of frass under reduced CO₂, which we used as a proxy for plant consumption, suggests that herbivores were deterred from feeding; thus, they would have not been able to access these N resources.

Understanding the diminishing levels of Si-based plant defences against herbivores under future elevated atmospheric CO₂ concentrations has received limited attention. Previous studies have shown that elevated CO₂ concentrations decrease Si accumulation in different Poaceae genera (e.g., grass species and wheat) (Johnson and Hartley, 2018; Biru et al., 2022; Johnson et al., 2022), and this was associated with reduced Si defences while increasing herbivore performance; however, this effect is not always reproducible (Frew et al., 2017). The impact of elevated CO₂ on Si defences reflects that plants switch from Si defences to C-based defences due to higher C availability under this scenario (Johnson and Hartley, 2018; Johnson et al., 2022). Although not examined in the context of Si defences, previous studies have demonstrated that elevated atmospheric CO₂ increased consumption and growth rate of the generalist (Pseudaletia unipuncta) and specialist (Spodoptera frugiperda) insect herbivores when fed on C₃ grass relative to lower atmospheric CO₂ conditions (Barbehenn et al., 2004). Johnson et al. (2020) also reported that H. armigera RGR increased under elevated CO₂ as a result of lower plant defence signalling and minimal reductions in the nutritional quality of lucerne (Medicago sativa).