Elevated Atmospheric CO2 Triggers Compensatory Feeding by Root Herbivores on a C3 but Not a C4 Grass

Predicted increases in atmospheric carbon dioxide (CO2) concentrations often reduce nutritional quality for herbivores by increasing the C∶N ratio of plant tissue. This frequently triggers compensatory feeding by aboveground herbivores, whereby they consume more shoot material in an attempt to meet their nutritional needs. Little, however, is known about how root herbivores respond to such changes. Grasslands are particularly vulnerable to root herbivores, which can collectively exceed the mass of mammals grazing aboveground. Here we provide novel evidence for compensatory feeding by a grass root herbivore, Sericesthis nigrolineata, under elevated atmospheric CO2 (600 µmol mol−1) on a C3 (Microlaena stipoides) but not a C4 (Cymbopogon refractus) grass species. At ambient CO2 (400 µmol mol−1) M. stipoides roots were 44% higher in nitrogen (N) and 7% lower in carbon (C) concentrations than C. refractus, with insects performing better on M. stipoides. Elevated CO2 decreased N and increased C∶N in M. stipoides roots, but had no impact on C. refractus roots. Root-feeders displayed compensatory feeding on M. stipoides at elevated CO2, consuming 118% more tissue than at ambient atmospheric CO2. Despite this, root feeder biomass remained depressed by 24%. These results suggest that compensatory feeding under elevated atmospheric CO2 may make some grass species particularly vulnerable to attack, potentially leading to future shifts in the community composition of grasslands.


Introduction
The largest annual increase in global atmospheric CO 2 emissions in the last 50 years occurred during 2010 [1]. Such increases will impact on ecological communities and the species interactions within them. For example, it is widely observed that elevated atmospheric CO 2 concentrations (eCO 2 ) reduces the nutritional quality of plants for herbivores [2]. A meta-analysis of over 100 published studies demonstrated that while both carbon (C) and nitrogen (N) increased in both roots and shoots, C increased at an accelerated rate relative to N and led to an average increase in C:N ratios of 11%, effectively reducing nitrogen concentrations in both roots and shoots [3], either by dilution or reallocation [4]. Since N is the limiting factor in most herbivore diets [5], herbivores may respond to this decline in host quality by compensatory feeding, whereby the herbivore eats more plant biomass in an attempt to acquire adequate nutrition [4,6]. Indeed, Stiling and Cornelissen's [2] meta-analysis reported that relative consumption by insect herbivores increased by 17% and total consumption by 19%, when feeding on plants under eCO 2 . Given that eCO 2 can cause similar increases in C:N ratios in the roots as in the shoots [3], it is surprising that feeding responses of root herbivores to such changes in chemistry have not been examined [7]. To our knowledge, only four studies have investigated the effects of eCO 2 on root feeding insects [8][9][10][11] and none have investigated this for grasses.
Root herbivores are major components of many ecosystems, having the capacity to shape the community structures of other herbivores and plant communities [12,13]. Grassland systems can be particularly vulnerable to root herbivores [14]. For instance, in some pasture systems it is not uncommon for the collective biomass of root herbivores to exceed that of grazing mammals aboveground [15]. In addition to covering over 40% of the planet's land surface area [16], grasslands are responsible for storing over one third of global terrestrial carbon stocks [17]. Grasslands often comprise of C 3 and C 4 grass species; C 3 grasses are usually superior hosts for herbivores compared with C 4 grasses, but they also are more strongly affected by elevated CO 2 , generally showing greater increases in C and reductions in N concentrations than C 4 plants [18,19]. This occurs because Rubisco, the initial carboxylating enzyme to facilitate the assimilation of CO 2 into carbohydrates operates below its maximum capacity at current CO 2 concentrations in C 3 plants, so has the greater capacity to respond to eCO 2 [4,20]. Because of this, C 3 plants may be disproportionately subject to compensatory feeding under eCO 2 . This has rarely been tested for aboveground herbivores (e.g. [21,22]), and never, to our knowledge, for root herbivores.
This study characterised how eCO 2 affected a C 3 (Microlaena stipoides) and a C 4 (Cymbopogon refractus) grass species, and how any changes in grass traits affected the feeding behaviour and performance of a root feeding insect, the scarab Sericesthis nigrolineata Boisduval (Coleoptera: Scarabaeidae). We hypothesised that: (1) M. stipoides would be nutritionally superior (higher N, lower C:N) than C. refractus under ambient CO 2 (aCO 2 ), (2) under eCO 2 , both grasses would become inferior hosts for the root herbivore due to lower N and higher C:N, and this effect would be more pronounced for M. stipoides than C. refractus; (3) under eCO 2 , insects would consume more root tissue through compensatory feeding, with the biggest increase on M. stipoides and (4) scarab performance (body mass) would better on M. stipoides than C. refractus, but would decline at eCO 2 .

Chambers
Six glasshouse chambers, three maintained at aCO 2 of 400 mmol mol 21 and the other three at eCO 2 (600 mmol mol 21 ), were used. These chambers (3 m65 m63 m; width6length6 height) with UV transparent plexiglass (6 mm thick) walls and roof were naturally lit throughout the experiment. Daytime air temperature was regulated to reach a midday peak of 24uC and fall to 21uC at midnight (64uC) at night time. Humidity was controlled at 60% (66%). CO 2 levels were controlled via the monitoring and control system PlantVisorPRO (Carel Industries, Padova, Italy). Briefly, CO 2 levels within each chamber were monitored by a CO 2 probe (GMP222, Vaisala, Vantaa, Finland), with CO 2 (food grade, AirLiquide, Australia) injected from pressurized cylinders through solenoid valves. Before entering a chamber, CO 2 was passed through a Purafils column to eliminate possible ethylene contamination.

Experimental procedure
Grasses were planted in 90 pots (100 mm diameter) containing 850 g of air-dried and sieved (1 mm) soil, which was loamy-sand with low (0.7%) organic matter (full details given in [23]). These were randomly assigned to the six climate chambers (15 in each) and watered daily to maintain soil water content at 15%, which was verified with a two-rod moisture probe (Hydrosense, Campbell Scientific, Australia). No nutritional supplement was provided. After 10 weeks, five of the plants for each species were selected and shoots and roots were separated, oven dried (40uC) and weighed. For the remaining 10 plants, a single seedling comprising small section of roots and grass blades (c. 2-3 g fresh mass) was teased apart and transferred into bioassay cages ( Fig. 1) constructed from 90 mm Petri dishes filled with soil (details as above). Cages were a variation of similar bioassay cages used for measuring root damage by root feeding insects in other studies [24,25]. One side had an aperture, through which the grass blades were left exposed. Dishes were wrapped in tinfoil and stored vertically in the chambers. Moisture was maintained by devlivering 2-3 ml water daily via the aperture. After 3 d, a single second instar larva was weighed and placed inside half of the cages, selected at random. Insects were from an established culture at UWS previously obtained from a site containing a range of C 3 and C 4 grasses [26]. After 7 d, the larva was removed and re-weighed. Roots were snap-frozen, freeze-dried and weighed. Material was milled and analysed for C and N concentrations using a LECO TruSpecH CHN analyser.

Statistical analysis
Analysis of variance (ANOVA) tests, in which chamber (and hence the three replicates of CO 2 treatment) were included as block terms to avoid pseudo-replication, were used. Plant biomass,  root consumption and larval mass was analysed with a two-way ANOVA (grass species and CO 2 , with an interaction of each term) with initial larval mass included as a covariate in the latter case. Chemistry was analysed with three-way ANOVAs (grass species, CO 2 and insect presence, with interactions of each term). Differences between treatments were determined using least square mean tests when significant interactions between CO 2 and grass species existed. Unless indicated otherwise all analysis was conducted on untransformed data using Genstat (version 14, VSN International, UK).

Plant responses
Plant biomass was unaffected by eCO 2 , for either grass species, although C. refractus plants were significantly bigger than M. stipoides largely due to higher shoot mass ( Table 1). Concentrations of root C were higher in C. refractus ( Fig. 2A) than M. stipoides (Fig. 2B), but largely unaffected by other variables (Table 2). In contrast, root N concentrations were higher in M. stipoides than C. refractus (Fig. 2 C-D), but in this case there was also a significant interactive effect of CO 2 and grass species (Table 2). In particular,  (Fig. 2C-D). Roots of C. refractus had a higher C:N ratio than M. stipoides (Fig. 2E-F). Again there was a significant interaction between eCO 2 and grass type (Table 1); eCO 2 caused an increase in root C:N in M. stipoides (Fig. 2F), but not in C. refractus (Fig. 2E).
Insect responses CO 2 did not affect root consumption overall (Fig. 3A) but consumption rose significantly on M. stipoides under eCO 2 . No difference in root consumption was seen on C. refractus under either CO 2 treatment (Fig. 3A). The final mass of larvae was higher on M. stipoides than C. refractus at aCO 2 , but under eCO 2 conditions larval mass was reduced to levels seen for those feeding on C. refractus (Fig. 3B).

Discussion
This study demonstrates that eCO 2 negatively affects a grass root herbivore when feeding on M. stipoides, a C 3 grass, but not on C. refractus, a C 4 grass. Elevated CO 2 caused bigger reductions in M. stipoides quality than in C. refractus, lowering N concentrations and increasing the C:N ratio and root consumption by insects. This increase in herbivore damage may make M. stipoides more susceptible to herbivory under predicted climate change than other competing species, with consequences for the composition of grassland systems. As hypothesised, we found that M. stipoides was a better quality host that C. refractus, and that eCO 2 had a bigger impact on M. stipoides. Our findings also matched our predictions that herbivore performance on M. stipoides would be worse under eCO 2 and compensatory feeding would take place, but we did not find that increased feeding compensated for this decrease in performance.
C 3 and C 4 grasses differ in physiological, anatomical and chemical traits which are thought to make C 3 grasses more susceptible to herbivory than C 4 grasses, giving rise to the C 3 -C 4 hypothesis [27]. This states that herbivores should select and perform better on C 3 rather than C 4 plants because they find them easier and more nutritious to consume. Compared to C 3 grasses, C 4 grasses tend to be lower in protein, which is also less accessible to herbivores as it is stored in bundle sheath cells, and possess higher levels of structural carbohydrates, making them less suitable host plants for folivores [27,28]. The lower demands and uptake of N by C 4 grasses compared to C 3 grasses may similarly result in lower root N concentrations in C 4 grasses and make them less  favourable for root herbivores [26]. Under aCO 2 , M. stipoides represented a superior host for root herbivores than the C 4 grass C. refractus, in line with the C 3 -C 4 hypothesis [27]. While tests with further species are needed, these findings provide some initial support for the C 3 -C 4 hypothesis potentially operating belowground as well as aboveground.
Despite evidence for compensatory feeding on the M. stipoides at eCO 2 , insect performance remained depressed at eCO 2 , suggesting that increased levels of herbivory were not enough to compensate for reduced plant quality. Similar effects occur for aboveground herbivores; Stiling and Cornelissen [2] concluded that most insect herbivores were generally unable to redress the problem of reduced food quality, and their abundance typically fell by 21% under eCO 2 . Compensatory feeding imposes extra energy requirements [6] which may be particularly demanding for soildwelling herbivores that have to physically burrow through the soil to access new root tissue [29]. Further work is needed to understand mechanisms of compensatory feeding for belowground herbivores, but the fact that they could not adequately compensate for deterioration in food quality suggests that energy constraints and thresholds for host plant quality may play a role. For example, the costs and benefits associated with compensatory feeding are likely to vary with both host quality and the impact of changes in quality on herbivore development and may only be possible above a quality threshold [30].
Compensatory feeding on C 3 grasses might be particularly damaging since we observed no significant increases in plant biomass in response to eCO 2 . In their review, Hovenden and Williams [31] also note that Australian grasses are generally unresponsive to eCO 2 in terms of growth, so they may be prone to higher herbivory levels without the advantage of enhanced growth rates seen in many other plants under eCO 2 . Our observations suggest that eCO 2 may contribute to compositional changes in grass communities if C 3 grasses are disproportionately damaged by root herbivores.