Effects of Elevated CO2 on Litter Chemistry and Subsequent Invertebrate Detritivore Feeding Responses

Elevated atmospheric CO2 can change foliar tissue chemistry. This alters leaf litter palatability to macroinvertebrate detritivores with consequences for decomposition, nutrient turnover, and food-web structure. Currently there is no consensus on the link between CO2 enrichment, litter chemistry, and macroinvertebrate-mediated leaf decomposition. To identify any unifying mechanisms, we presented eight invertebrate species from aquatic and terrestrial ecosystems with litter from Alnus glutinosa (common alder) or Betula pendula (silver birch) trees propagated under ambient (380 ppm) or elevated (ambient +200 ppm) CO2 concentrations. Alder litter was largely unaffected by CO2 enrichment, but birch litter from leaves grown under elevated CO2 had reduced nitrogen concentrations and greater C/N ratios. Invertebrates were provided individually with either (i) two litter discs, one of each CO2 treatment (‘choice’), or (ii) one litter disc of each CO2 treatment alone (‘no-choice’). Consumption was recorded. Only Odontocerum albicorne showed a feeding preference in the choice test, consuming more ambient- than elevated-CO2 birch litter. Species’ responses to alder were highly idiosyncratic in the no-choice test: Gammarus pulex and O. albicorne consumed more elevated-CO2 than ambient-CO2 litter, indicating compensatory feeding, while Oniscus asellus consumed more of the ambient-CO2 litter. No species responded to CO2 treatment when fed birch litter. Overall, these results show how elevated atmospheric CO2 can alter litter chemistry, affecting invertebrate feeding behaviour in species-specific ways. The data highlight the need for greater species-level information when predicting changes to detrital processing–a key ecosystem function–under atmospheric change.


Introduction
Global concentrations of atmospheric carbon dioxide (CO 2 ) could more than double by 2100 [1]. Typically, CO 2 enrichment leads to increased plant photosynthesis, resulting in greater biomass and production [2]. Plant tissue chemistry is typically modified, with decreasing nitrogen concentrations and increasing carbon-nitrogen (C/N) ratios affecting herbivore life-history and feeding responses [3].
Approximately 90% of primary production in forest ecosystems escapes herbivory and forms detritus [4], providing a crucial energy pool that underpins the trophic structure of soils and adjacent freshwaters [5]. The effect of elevated CO 2 on the chemical composition of green foliar tissues reduces its palatability to detritivores when it falls as litter [6]. In particular, elevated CO 2 can reduce litter resource quality by decreasing litter nitrogen content [7,8], subsequently increasing C/N ratios [9,10]. Increases in structural [6,8,9] and defensive [10,11] compounds have also been reported, along with both increases and decreases in phosphorus concentrations [12,13]. The potential for rising CO 2 concentrations to alter litter chemical composition is established, but the consequences for invertebrate-mediated decompositionan important ecosystem function -remain unclear [14].
Detritivorous macroinvertebrates are functionally important in detritus-based ecosystems, as they are responsible for both comminution and consumption of litter, releasing nutrients for other organisms, such as saprophagous fungi [15,16]. To maintain optimal body nutrient concentrations, theoretical predictions and empirical evidence suggest that invertebrates can increase feeding rates of reduced-quality material (e.g. [17,18]), a process known as 'compensatory feeding' (as defined by [19]). Despite this, poor quality litter has also been shown to increase handling times [20], while reducing nutrient assimilation, slowing development rates, and increasing mortality [6,21]. These conflicting responses have resulted from studies focusing on a small number of species (e.g. [13,18]), which also fail to incorporate aquatic and terrestrial invertebrates, despite differences in detrital accumulation and energy flow between these habitats [22]. A broad-scale study incorporating a range of invertebrate species from different habitats is essential to identify the unifying mechanisms that govern invertebrate feeding responses to elevated-CO 2 litter.
We investigated the feeding preferences and consumption rates of eight detritivorous macroinvertebrate species presented with Alnus glutinosa (Linnaeus) Gaertner (common alder) and Betula pendula Roth (silver birch) leaf litter produced under ambient and elevated atmospheric CO 2 . We tested the hypotheses that: (1) CO 2 enrichment will reduce leaf chemical quality and, given nitrogenfixing ability in alder, responses will differ by tree species; (2) when presented with a choice between ambient and elevated CO 2 litter, invertebrates will prefer ambient material due to its higher quality; (3) when given litter of one CO 2 treatment only, consumption of elevated-CO 2 litter will be greater, to compensate for its reduced quality.

Leaf Litter Preparation
Alder and birch litters were produced at the BangorFACE facility, Bangor, UK [23] (Fig. 1). Trees were grown in eight identical plots (four ambient-CO 2 and four elevated-CO 2 ) to minimise infrastructure-induced artefacts. CO 2 enrichment was carried out using high velocity pure CO 2 injection, controlled using equipment and software modified from EuroFACE [24]. Elevated CO 2 concentrations, measured at 1 min intervals, were within 30% deviation from the pre-set target concentration of 580 ppm CO 2 (ambient +200 ppm) for 75-79% of the photosynthetically-active period (daylight hours from budburst until leaf abscission) of 2005-2008. Vertical profiles of CO 2 concentration measured at 50 cm intervals through the canopy showed a maximum difference of +7% from reference values obtained at the top of the canopy [23]. From the beginning of leaf senescence, fallen leaf litter was collected weekly until all leaves had abscised (October to December). Litter within each CO 2 treatment was homogenised and air-dried.
Initial chemical leaching and microbial colonisation of litter ('conditioning') are crucial steps in making litter palatable to detritivorous macroinvertebrates [25,26]. Prior to the start of the experiment, litter was conditioned in fine mesh bags (100 mm to permit microorganisms only) placed in plastic containers (29629610 cm; Fig. 1). For each tree species 6CO 2 treatment combination, one bag was placed in aerated stream water that was inoculated with stream-collected litter of mixed-species origin ('aquatic conditioning'); a second bag per tree species 6CO 2 treatment combination was inserted between field-collected soil and mixed deciduous leaf litter ('terrestrial conditioning'). Containers were maintained at 1161uC with a 12:12 h light-dark cycle and terrestrial containers were sprayed with deionised water every three days to maintain humidity (,50%). These conditions were selected to represent natural conditioning processes in aquatic and terrestrial habitats in a controlled manner. After two weeks, leaf discs were cut using a 9 mm diameter cork-borer (avoiding the mid-vein), which were air-dried and weighed (60.1 mg) prior to experimental use.
Litter samples allocated to chemical analyses ( Fig. 1) were stored at -80uC before being oven-dried (50uC for 24 h) and ground into powder (120 s, 50 beats s -1 ; Pulverisette 23 ball mill, Fritsch GmbH, Idar-Oberstein, Germany). Each sample was composed of litter from three separate leaves. For carbon, nitrogen and phosphorus analyses, five samples were processed per tree 6 CO 2 treatment 6 conditioning type combination; for lignin analysis, four samples were used. The percentage leaf dry mass (% leaf DM) of carbon and nitrogen, and the carbon-nitrogen (C/N) ratio, were determined by flash combustion and chromatographic separation of ,1.5 mg leaf powder using an elemental analyser (Elemental Combustion System 4010 CHNS-O Analyzer, Costech Analytical Technologies, Inc., Milan, Italy), calibrated against a standard (C 26 H 26 N 2 O 2 S). Phosphorus concentrations (% leaf DM) were quantified using X-ray fluorescence (see [27] for detailed methodology). The percentage Acetyl-Bromide-Soluble Lignin (% ABSL) was determined following the acetyl bromide spectrophotometric method [28]. Lignin-nitrogen (lignin/N) ratios were calculated for each tree species 6 CO 2 treatment 6 conditioning treatment combination.

Invertebrates
Eight macroinvertebrate species were selected for study (Table 1), representing a taxonomic range of litter consumers found in temperate forest habitats [29,30]. Aquatic species were collected from streams in the Brecon Beacons National Park, South Wales, UK (51u509530N, 3u229160W and 51u509550N, 3u339430W) and Roath Park, Cardiff, UK (51u309000N, Figure 1. Overview of the experimental approach. Litter was produced under ambient-and elevated-CO 2 atmospheres at BangorFACE, UK. Half of the litter from each CO 2 treatment was conditioned aquatically and half terrestrially. Chemical analyses of the conditioned litter were undertaken, and litter discs were presented to aquatic and terrestrial invertebrates in choice and no-choice tests. Only one tree and one invertebrate species have been shown for clarity. Not to scale. doi:10.1371/journal.pone.0086246.g001 3u109100W); terrestrial species were collected from soil-litter interfaces in Bute Park, Cardiff, UK (51u489490N, 3u189240W). The National Park Authority granted general permission to access sites on common land in the Brecon Beacons National Park, South Wales, UK. Cardiff Council granted permission for access to sites in Cardiff, UK. No endangered or protected species were involved in collections from the field. All individuals were adults, apart from larval Odontocerum albicorne and Sericostoma personatum caddisflies. Individuals from within each species were selected for size similarity. Prior to experimental use, invertebrates were maintained for at least four weeks in single-species containers (1161uC, 12:12 h light-dark cycle) and were fed Fagus sylvatica Linnaeus (common beech) litter conditioned as for experimental litter, preventing habituation to experimental alder and birch litter. Feeding was ceased two days prior to the experiments to allow for gut clearance.

Experimental Arenas
All experiments were conducted in 11616.563.5 cm lidded plastic arenas (Cater For You Ltd, High Wycombe, UK) lined with compacted sterilised aquarium gravel (Unipac, Northampton, UK) and were maintained at 1161uC with a 12:12 h light-dark cycle. Aquatic microcosms were filled with 400 ml of filtered (100 mm mesh) stream water (circumneutral pH; collected from 51u509530N, 3u229160W) and aerated through a pipette tip (200 ml Greiner Bio-One) attached to an air-line. Terrestrial microcosms were sprayed with deionised water every three days to maintain moisture content and humidity (,50%). All arenas were uniquely labeled ('microcosm ID'). These standardised conditions were chosen to mimic natural habitats, while minimising the availability of supplementary organic material that could act as a confounding resource during the feeding trials.
For litter of each tree species, detritivores were presented with: (i) a choice between ambient-and elevated-CO 2 material, to provide a direct comparison of detritivore preferences, and (ii) a no-choice situation with each CO 2 treatment presented on its own, approximating litter consumption in current (ambient-CO 2 ) and future (elevated-CO 2 ) atmospheric conditions (Fig. 1). In each experiment, ten microcosms were set up for each invertebrate and tree species combination (n = 160). A single invertebrate was added to each arena and was placed in the end opposite the airline in aquatic arenas and equidistant to both discs in the choice test. In the choice test, one disc of each CO 2 treatment was pinned to the centre of the arena, 4 cm apart. Discs were replenished when at least 50% of the existing disc had been consumed. In the no-choice test, half of the microcosms contained one ambient-CO 2 disc and the other half one elevated-CO 2 disc, pinned to the centre of the arena. Both experiments ended after 14 days, or when five (50%) of the individuals of a specific species consumed at least 50% of one disc (choice experiment only). For each invertebrate, the total mass of litter consumed was calculated (60.1 mg). For choice experiment data, this value was divided by the number of days over which the test had taken place.
Additionally, control microcosms were set up to ensure that differences in mass loss between CO 2 treatments were due to invertebrate activity alone. For each experiment, ten microcosms were set up for each habitat type 6 tree species combination. Controls for the choice test each contained one disc of each CO 2 treatment; half of the no-choice control microcosms contained one ambient-CO 2 disc and the other half contained one elevated-CO 2 disc. Leaf discs were air-dried and weighed (60.1 mg) after 14 days and their total mass loss calculated.

Data Analysis
Statistical analyses were performed separately for alder and birch litter using R version 3.0.1 [31]. Data available from http:// dx.doi.org/10.6084/m9.figshare.791634. were checked for normality and homogeneity of variance following Crawley [32]; response variables were transformed using Box-Cox power transformations when assumptions were not met (car package [33]). Significance was set at a = 0.05 for all analyses.
Two-way analysis of variance (ANOVA) was used to test the main and interactive effects of CO 2 treatment and microcosm type on each chemical variable (carbon, nitrogen, phosphorus and lignin concentrations, and C/N ratio). Planned contrasts (lsmeans package [34]) were used to compare the effects of CO 2 treatments for each conditioning treatment.
The main and interactive effects of CO 2 treatment and microcosm type were tested on the mass loss of control discs. Linear mixed-effects models were used to analyse choice control data (nlme package [35]), where non-independence of discs sharing the same microcosm was accounted for by including microcosm ID as a random term. The same fixed terms were used to analyse control data from the no-choice test using two-way ANOVA.
In the choice test, litter consumption per day was analysed using linear mixed-effects models (nlme package [35]) with the main and interactive effects of CO 2 treatment and invertebrate species as fixed effects and microcosm ID as a random effect. Planned contrasts were performed to compare consumption of ambientand elevated-CO 2 discs within (i) each invertebrate species, and (ii) invertebrate species grouped by habitat of origin (contrast package [36]).
In the no-choice test, the main and interactive effects of CO 2 treatment and invertebrate species on litter consumption were tested using two-way ANOVA. Planned contrasts were performed to test the effects of CO 2 treatment on disc consumption within (i) each invertebrate species (lsmeans package [34]) and (ii) invertebrate species grouped by habitat of origin (gmodels package [37]).

Invertebrate Responses
For both tree species in the choice and no-choice control arenas, disc mass loss in the absence of invertebrates was unaffected by CO 2 treatment and conditioning type (P.0.05). Litter mass loss in the presence of invertebrates was therefore assumed to be a result of invertebrate feeding alone.

Discussion
Elevated atmospheric CO 2 and microbial conditioning type modified leaf litter chemistry, though effects differed between tree species (supporting Hypothesis 1). Individual invertebrate species varied in their responses, suggesting that caution has to be taken when extrapolating general trends from single-species studies.
Elevated atmospheric CO 2 reduced birch litter quality: the concentration of nitrogen decreased and the C/N ratio increased, regardless of conditioning type. Most species did not respond to this change; O. albicorne was the only species with behaviour that supported Hypothesis 2, showing a strong preference for ambient-CO 2 litter. Prior work supports this response: Ferreira et al. [13] showed that low C/N ratios reduced birch litter consumption by the caddisfly Sericostoma vittatum Rambur, while Cotrufo et al. [17] found that the woodlouse P. scaber preferred high quality (lower C/ N ratio and lignin concentration) Fraxinus excelsior Linnaeus litter grown under ambient CO 2 . Alder litter showed negligible chemical change as a result of elevated CO 2 , perhaps due to symbiosis with nitrogen-fixing bacteria that help maintain nutrient supplies [38]. Unexpectedly, a slight increase in quality (increased Table 2. ANOVA summary table of main and interactive effects of CO 2 treatment (CO 2 ) and conditioning type (CT) on litter chemistry. nitrogen concentration) under elevated CO 2 occurred when alder litter was conditioned terrestrially, but this did not result in any feeding preferences. Effects of conditioning type on litter chemistry may have occurred due to differences in chemical leaching and microorganism activity between aquatic and terrestrial environments [39]. Our data indicate that CO 2 enrichment will affect litter palatability to macroinvertebrate detritivores as a result of chemical change, though these effects will be plant and invertebrate species-specific. In the no-choice test, invertebrates were expected to compensate for low-quality litter by increasing consumption relative to high-quality litter. In contrast to this expectation, compensatory feeding was not observed in either tree species. There was no clear pattern for alder; invertebrate responses were highly idiosyncratic, with O. asellus being the only species to consume more of the low-  quality resource (terrestrially-conditioned alder litter contained lower nitrogen when grown under ambient-CO 2 ). Hä ttenschwiler et al. [18] detected a similar compensatory response for O. asellus and another woodlouse, P. scaber: higher consumption rates were recorded on low-quality, CO 2 -enriched F. sylvatica litter (low nitrogen concentration, high C/N ratio). The current study showed that G. pulex and O. albicorne consumed more elevated-CO 2 than ambient-CO 2 alder, despite no observed chemical differences. It is possible that elevated CO 2 reduced litter palatability by altering chemical constituents that were not quantified here, such as secondary metabolites. For example, phenolics and tannins have been shown to be affected by CO 2 levels [40]. Birch litter responses appeared less idiosyncratic, with no individual species increasing consumption of elevated-CO 2 litter. These results suggest that litter species identity determines the predictability of invertebrate feeding responses, but that compensatory feeding is not a unifying trend amongst detritivorous macroinvertebrates. Feeding rates may have varied due to increased handling times associated with low quality birch litter (e.g. [20]), or because of differences in species' body chemistry and their ability to cope with elemental imbalances with CO 2 -enriched resources [41,42]. Heterotrophs, such as the detritivores in our study, tend to maintain constant body elemental composition [43] and may alter feeding behaviour to achieve optimum chemical balance. Our results show that individual invertebrate species rarely demonstrated significant responses to CO 2 treatments in either test. This suggests that although individual species responses appear idiosyncratic, when considered as a whole, the invertebrate community generally shows consistent and predictable behavioural and functional responses to litter chemical changes induced by elevated CO 2 .
Altered consumption of litter by macroinvertebrates will affect energy release from detritus, in turn affecting secondary production, and food-web structure and functioning [5]. Specifically, on the basis of invertebrate responses in our study, mineralisation of carbon and nutrients could slow down in forests dominated by birch or other tree species with similar chemistry. This is reinforced by our observations of high lignin/N and C/N ratios of elevated-CO 2 birch litter, which are predictors for slow decomposition rates [44]. Conversely, stands containing a lot of alder, or other species with lower C/N ratios, may show little response in terms of detrital processing and nutrient turnover. Differences between tree species make it difficult to predict overall decomposition rates, a task made more difficult by the prevalence of litter mixtures in temperate deciduous forests, which tend to exhibit non-additive decay [45].
Changes to litter quality as a result of elevated CO 2 may also affect invertebrate community composition, a potentially important determinant of decomposition rates [19]. This could be caused by changes to food selection [46] and increased patchiness of resource quality in litter mixtures on the forest floor [47]. Differential changes to feeding rates may alter competitive dynamics between invertebrate species, with advantages for species whose dietary breadth extends beyond leaf litter, such as G. pulex and S. personatum [48,49].
Our study provides, to date, the broadest assessment of detritivorous invertebrate species' feeding responses to CO 2enriched litter, improving our mechanistic understanding of a key ecosystem process in temperate woodland ecosystems. Future elevations of atmospheric CO 2 are predicted to affect the breakdown of detritus indirectly by reducing leaf litter quality for macroinvertebrate detritivores. The study highlights that this process is highly tree species-specific, and there will be strong responses in some forest stands and minimal effects in others. Identifying the mechanisms governing such ecosystem variation in functional responses to climate change is essential if we are to predict the consequences of elevated CO 2 for forest carbon dynamics and nutrient cycling at regional and landscape-scales.