Litters of photosynthetically divergent grasses exhibit differential metabolic responses to warming and elevated CO2

Climatic stress such as warming would alter physiological pathways in plants leading to changes in tissue chemistry. Elevated CO2 could partly mitigate warming induced moisture stress, and the degree of this mitigation may vary with plant functional types. We studied the composition of structural and non-structural metabolites in senesced tissues of Bouteloua gracilis (C4) and Pascopyrum smithii (C3) at the Prairie Heating and CO2 Enrichment experiment, Wyoming, USA. We hypothesized that P. smithii and B. gracilis would respond to unfavorable global change factors by producing structural metabolites and osmoregulatory compounds that are necessary to combat stress. However, due to the inherent variation in the tolerance of their photosynthetic pathways to warming and CO2, we hypothesized that these species will exhibit differential response under different combinations of warming and CO2 conditions. Due to a lower thermo-tolerance of the C4 photosynthesis we expected B. gracilis to exhibit a greater metabolic response under warming with ambient CO2 (cT) and P. smithii to exhibit a similar response under warming combined with elevated CO2 (CT). Our hypothesis was supported by the differential response of structural compounds in these two species, where cT increased the content of lignin and cuticular-matrix in B. gracilis. In P. smithii a similar response was observed in plants exposed to CT, possibly due to the partial alleviation of moisture stress. With warming, the total cell-wall bound phenolic acids that cross link polysaccharides to lignins increased in B. gracilis and decreased in P. smithii, indicating a potentially adaptive response of C4 pathway to warming alone. Similarly, in B. gracilis, extractable polar metabolites such as sugars and phenolic acids increased with the main effect of warming. Conversely, in P. smithii, only sugars showed a higher abundance in plants exposed to warming treatments indicating that warming alone might be metabolically too disruptive for the C3 photosynthetic pathway. Here we show for the first time, that along with traditionally probed extractable metabolites, warming and elevated CO2 differentially influence the structural metabolites in litters of photosynthetically divergent grass species. If these unique metabolite responses occur in other species of similar functional types, this could potentially alter carbon cycling in grasslands due to the varying degradability of these litters.


INTRODUCTION
Responses of plants to changing abiotic and biotic environments could have cascading ecological implications. One such response is the alteration of metabolite profiles in plants exposed to climatic changes, as this response may potentially influence soil carbon storage by altering the chemistry of litter available for decomposition (Tharayil et al. 2011). Although it is well known that the stress adaptation might alter plant metabolic pathways (Dixon and Paiva 1995, Guy et al. 2008, Krasensky and Jonak 2012, most of the studies to date have primarily focused on capturing the compositional changes in metabolite-pools that are readily extractable, and less studied are the climate induced changes in non-extractable structural matrix of plant tissues that forms a major substrate for decomposition. The magnitude of stress response of plants would also depend on the combination of environmental conditions to which they are exposed to, and hence may vary with the physiological adaptation of plant functional types. To date, few studies have evaluated the combined effect of warming and elevated CO 2 on the structural and non-structural metabolites in senesced tissues of photosynthetically divergent species. Understanding these responses could provide a greater insight into potential plant species distributions and soil carbon cycling in a changing world. The perception and subsequent adaptation of plants to different climatic conditions could differ between species that vary in their photosynthetic carbon assimilation pathways. Generally, climates that are warmer and drier result in partial closure of stomata, which limit the growth of C3 species due to increased photorespiration caused by the lower partial pressure of CO 2 in the vicinity of RUBISCO (Sage and Kubien 2007). However, a concomitant increase in atmospheric CO 2 concentration could allow the RUBISCO to efficiently undertake the carboxylation of RuBP even under the partial closure of stomata, which could partly mitigate the detrimental effect of warmer drier climates (Pearcy andEhleringer 1984, Sage andKubien 2007) in C3 species. Conversely, C4 species are less responsive to elevated CO 2 due to the inherent physical separation of carbon assimilation and fixation that reduces the propensity of RUBISCO to undertake oxygenation reactions (Pearcy andEhleringer 1984, Sage andKubien 2007). Thus, generally C3 species are found to respond more positively to elevated CO 2 than C4 species with respect to both photosynthesis and biomass production (Smith et al. 1987, Tissue et al. 1995, Wand et al. 1999. For example, in a semi-arid grassland exposed to elevated CO 2 and increased soil moisture, the biomass production of C3 grasses increased under elevated CO 2 while C4 grasses increased their biomass only under increased soil moisture (Dijkstra et al. 2010).
The interactive effect of multiple climatic factors could elicit synergistic, antagonistic or additive plant responses (Dieleman et al. 2012). Although elevated CO 2 facilitates higher plant productivity, this plant response could be altered by accompanying warming or soil moisture stress. Heat stress in general decreases the benefits of elevated CO 2 on photosynthesis by enhancing leaf temperatures in both C3 and C4 species (Lara and Andreo 2011). However, since the stomatal conductance of C4 species is lower than that of C3 species at any given CO 2 concentration, the magnitude of this reduction in conductance is higher in C4 species, resulting in higher leaf temperatures , Lara and Andreo 2011. Thus, in general, C3 species exhibit a high thermotolerance of photosynthesis under elevated CO 2 (Sage et al. 1995), whereas in C4 species photosynthetic heat tolerance decreases under CO 2 enrichment , resulting in poor performance of C4 species under combinations of warming and elevated CO 2 . For example, a previous study has reported higher biomass production in C4 species when exposed elevated CO 2 combined with ambient temperature, whereas C3 species exhibited a similar response when elevated CO 2 was combined with warming (Hunt et al. 1996).
Grassland ecosystems, which occupy ;30% of the Earth's land area and contain 20% of the Earth's terrestrial soil carbon (FAO 2010, Craine et al. 2013, are dominated by C3 and/or C4 species that exhibit different photosynthetic pathways and capacities to thrive under future climates. Multi-factor climate experiments have shown that C4 grasses flourish under warming (Morgan et al. 2011). Although the overall responses of the C3 and C4 pathways to various global change factors have been well studied in terms of photosynthetic efficiency and biomass production (Crafts-Brandner and Salvucci 2002, Sage 2002, Lee 2011, remarkably little is known about how the C3 and C4 metabolic response translates to the chemical composition (structural and non-structural) of the senesced tissues, which in turn, could influence litter decomposition and soil carbon storage.
Studies on the effect of climatic stress on plant physiology have generally focused on the metabolite profiles of actively growing green tissues (Du et al. 2011, Rivas-Ubach et al. 2012, Yu et al. 2012. Although elucidating the changes in metabolites in green tissues aids in understanding the underlying physiological mechanisms through which plants adapt to stressful environments, by overlooking the nutrient resorption process that occurs during tissue senescence, such approaches provide little insight into the final composition of the litter that forms the substrate for microbial decomposition. The resorption efficiency for nitrogen and phosphorous compounds exceeds 50% in terrestrial plants (van Heerwaarden et al. 2003), and this remobilization is altered in plants experiencing stressful growing conditions (van Heerwaarden et al. 2003, Kobe et al. 2005. Thus, the metabolite composition of senesced tissues will be significantly different from that of green tissues and might be further altered under climatic stress. Our knowledge of the effects of multiple global change factors, especially warming and elevated CO 2 , on the chemical composition of plant litter is still rudimentary, which hinders our efforts to accurately predict the decomposition susceptibility of plant litter produced under future climates. We used two grass species, Pascopyrum smithii (C3) and Bouteloua gracilis (C4), collected from the Prairie Heating and CO 2 Enrichment (PHACE) experiment, Wyoming, USA to study the differential responses of the C3 and C4 pathways to elevated CO 2 and warming. These two species contribute to ;50% of the biomass production at PHACE and 90% of the C4 species at the study site is dominated by B. gracilis (Dijkstra et al. 2010, Morgan et al. 2011. The main objective of this study was to assess the differences in both structural and non-structural metabolite composition in senesced litters of P. smithii (C3) and B. gracilis (C4) that were exposed to a factorial combination of warming and elevated CO 2 . We hypothesized that the litters of both P. smithii and B. gracilis that are exposed to unfavorable global change conditions would be abundant in structural metabolites and osmoregulatory compounds that are necessary to combat stress. However, due to the variation in their photosynthetic pathways and differences in their tolerance to warming and CO 2 levels, we hypothesized that P. smithii and B. gracilis may show the above metabolic-responses under different combinations of warming and CO 2 enrichment. We predicted that warming alone would hamper any metabolic changes in P. smithii potentially due to the high moisture stress in this semiarid grassland; whereas warming accompanied by elevated CO 2 would lead to a greater metabolic response in P. smithii, potentially due to the partial alleviation of moisture stress. In contrast, because warming combined with elevated CO 2 leads to higher leaf temperatures in C4 species, we expected B. gracilis to exhibit a greater metabolic response to warming in the absence of elevated CO 2 . Thus, the metabolic responses of P. smithii and B. gracilis to warming and CO 2 conditions may be modulated by the intensity of the stress perceived by each species under different combinations of warming and CO 2 enrichment.

Site and experimental design
The plant litter examined in this study was obtained from the Prairie Heating and CO 2 Enrichment (PHACE) experiment located at Cheyenne, Wyoming, USA (latitude 41811 0 N, longitude 104854 0 W). The study site is a semiarid mixed grass prairie where 50% of the vegetation is dominated by the cool-season C3 grass Pascopyrum smithii (Rydb.) A. Love and the warm-season C4 grass Bouteloua gracilis (H.B.K) Lag. The above ground biomass production of B. gracilis and P. smithii at PHACE under ambient conditions ranged from 60 to 110 g m À2 and from 150 to 170 g m À3 , respectively (Morgan et al. 2011). The site exhibits a mean maximum temperature of 17.58C in July and a minimum of À2.58C in January, together with mean annual precipitation of 384 mm. The PHACE experimental plots were subjected to a factorial combination of two levels of CO 2 treatment (385 p.p.m.v. and elevated to 600 p.p.m.v.; abbreviated as c and C, respectively) and two levels of warming treatment (ambient and þ1.5/ 3.08C warmer in the day/night; abbreviated as t and T, respectively). There were five replicates for each treatment in a total of 20 circular plots with a diameter of 3.3 m. Infrared heaters were used to raise the canopy temperature to 1.58C above ambient temperature during the day and 38C above ambient temperature during the night. Carbon dioxide was applied using free air CO 2 enrichment (FACE) technology (Dijkstra et al. 2010, Morgan et al. 2011. In the semiarid PHACE experimental site, soil moisture content varied with different combinations of warming and elevated CO 2 . In general, soil moisture increased in elevated CO 2 with ambient temperature treatment while warming with ambient CO 2 treatment resulted in faster and severe soil drying (Morgan et al. 2011, Dijkstra et al. 2012). However, the soil moisture content of warming plus elevated CO 2 treatment was similar to that in the control as the warming induced loss of soil moisture was compensated by lower transpiration rate due to elevated CO 2 (Morgan et al. 2011). The volumetric water content of ambient treatment and CT treatments were 15.5% and 15.6% respectively. Compared to the ambient treatment, the average volumetric soil water content (SWC; 2007-2009) of elevated CO 2 plots increased by 12% whereas warming alone treatment reduced soil moisture content by 16% (Morgan et al. 2011). The productivity of C3 and C4 species in the different treatments also mirrored the changes in soil moisture content indicating that plants experienced moisture stress under different warming and elevated CO 2 treatments. The cool season, C3 species increased productivity (34%) in elevated CO 2 compared to the control, while warming favored C4 species (28% increase in productivity compared to control; Morgan et al. 2011).

Litter collection and processing
We collected senesced leaf litter of B. gracilis and P. smithii from the experimental plots during the fall of 2010. The litter was collected just after complete senescence while the leaves were still attached to the plant. The litter was collected as part of the biomass harvest done at the end of the growing season. The litter was air-dried (,358C) and about five grams from each plot were ground to a fine powder with a ball mill.
The samples were first extracted with aqueous methanol, and the polar metabolites such as amino acids, phenolic acids, organic acids, sugars and sugar alcohols (Table 1) in this extract were subjected to targeted metabolomics analysis to identify changes in the composition of nonstructural compounds in the senesced litter that are often highly susceptible to microbial degradation. The methanol extracted litter was further subjected to mild-alkaline hydrolysis using 1N NaOH to release the metabolites that were ester linked to the cell-wall, and the hydrolysates were analyzed for phenolic compounds (Table 2). We further used diffuse reflectance infra-red Fourier transform (DRIFT) spectroscopy to characterize the structural chemistry of both the non-extracted and methanol extracted tissues.

Litter chemistry analyses
Extractable polar metabolites.-Metabolic profiling of plants is a powerful tool that provides unbiased insight into plant tissue chemistry (Fiehn et al. 2000, Meyer et al. 2007, Tohge and Fernie 2010. The polar metabolites were extracted from the ground samples as per Lisec et al. (2006 ) and Kind et al. (2009), with slight modifications. Briefly, 50-100 mg samples were extracted with 1.5 ml of methanol containing ribitol (10 lg) as internal standard at 508C for 15 minutes in 2 ml-microcentrifuge tubes. The tubes were centrifuged (12,000g for 5 minutes), and the supernatants were transferred to 8-ml glass culture tubes containing 1 ml of water and cooled to 48C, after which the non-polar compounds from the extracts were partitioned with 500 ll of chloroform at 48C in a rotary shaker. Next, the tubes were centrifuged at 1,500 g for 10 minutes, and the top, aqueous methanol phase was drawn into glass vials. Subsamples (100-150 ll) were transferred to vials with glass inserts (250 ll) and dried completely under nitrogen. Five microliters of a fatty acid methyl ester standard mixture (C4-C30, even carbon) in hexane (100 lg ml À1 ) and 5 ll of d27-myristic acid in hexane (1 mg ml À1 ) were added to the vials, which were then dried completely under v www.esajournals.org nitrogen. The samples were further methoxylaminated via reaction with 20 ll of methoxylamine hydrochloride (20 mg ml À1 ) in pyridine at 408C for 90 minutes. Additionally, the acidic protons in the compounds were silylated via reaction with 80 lL of N-methyl-N-(trimethylsilyl) trifluoroacetamide (MSTFA) with 1% trimethylchlorosilane (TMCS) for 40 minutes at 408C. The samples were finally stored at 48C and were subjected to gas chromatography-mass spectrometry (GC-MS) analysis within 10 hrs of derivatization. The samples were analyzed using an Agilent 7980A GC system coupled with a 5975 C series mass detector. The separation of metabolites was achieved in a DB-5 MS capillary column (30 m length 3 0.25 mm internal diameter 3 0.20 lm film thickness) using split (1:10 and 1:100) injection (1 ll) with the following temperature program: 608C for 1 minute, followed by ramping at 108C per minute to 3158C, with a 10minute hold at 3158C prior to cool down. The carrier gas (He) was maintained at a constant pressure of 10.7 psi; the injection port and the MS interphase were maintained at 2708C; the MS quad temperature was maintained at 1508C; and the MS source temperature was set at 2408C. The electron multiplier was operated at a constant gain of 10 (EMV ¼ 1478 V), and the scanning range was set at 50-600 amu, achieving 2.66 scans sec À1 . The mass spectra were further processed using the Automatic Mass Spectral Deconvolution and Identification System (AM-DIS, v2.71, NIST) with the following deconvolution parameters: match factor, 75%; resolution, high; sensitivity, medium; shape requirements, medium. The compounds were positively identified based on an in-house retention index mass spectral library supplemented with Fiehn Lib (Agilent Technologies, Wilmington, Delaware, USA, G1676AA), which contains the retention indices and mass fragment information for 768 plant metabolites, and the NIST11 mass spectral library. All positive matches were confirmed by manual curation, and the integrated area with reference to internal standard was used for further statistical analysis. The extractable polar metabolites were grouped into amino acids, organic acids, phenolic acids, sugars and sugar alcohols for statistical analyses.
Cell wall bound phenolics (ester-bound phenolics).-This extraction procedure removes the ester-bound phenolics from cell wall tissue via mild alkaline hydrolysis (Martens 2002). Briefly, 200 mg of leaf litter that was prior extracted with methanol (used for metabolomics analysis) was combined with 6 ml of 1 M NaOH for alkaline hydrolysis and incubated at 258 6 28C for 15 hours on an orbital shaker at 26 rpm. The supernatant was incubated on a heating block for two hours at 908C to release the conjugated phenolics. Next, the solution was acidified (pH , 1.5) using 50% HCl and centrifuged at 2,000 rpm for 4 minutes. The supernatant was volumetrically transferred to glass tubes, and the phenolic compounds in the solution were recovered through liquid-liquid partitioning with 2 ml of ethyl acetate on an orbital shaker at 48C. The percent recovery of all phenolic compounds v www.esajournals.org using this method was .75% . A subsample of the ethyl acetate fraction was evaporated to dryness under N 2 and redissolved in 50:50 (v/v) methanol:water for analysis via high performance liquid chromatography (HPLC). HPLC analysis of the ester fractions.-Phenolics were separated in an Onyx C18 column (monolithic silica, 130A 0 ; 100 mm; 4.6 mm I.D.; Phenomenex, Torrance, CA) via binary gradient elution using 5% acetonitrile containing 0.2% acetic acid (mobile phase A) and 50% acetonitrile (mobile phase B), with a linear increase in the strength of the mobile phase from 100% A at 0 min to 76% at 17 min and a flow rate of 0.8 ml min À1 . These settings provided a minimum peak resolution (Rs) of 1.8. The limit of detection and quantitation was defined as a signal-to-noise ratio greater than 8 and 25, respectively. The reported values are based on the peak area at 272 nm. The compounds in the extracts were identified and quantified by comparison of the obtained retention times and UV spectra to those of an in-house library of 42 plant phenolic compounds .
DRIFT spectroscopy.-Senesced litter that was not subjected to solvent extraction and the residual litter samples remaining after methanol extraction were used in the DRIFT analyses. The samples were dried at 508C to remove any moisture and further powdered to ,10 lm particle size. The DRIFT spectra of the litter samples were collected in transmission mode using a spectrometer (Perkin-Elmer Spectrum One DRIFT) equipped with a deuterated triglycine sulfate detector. The finely powdered litter samples were mixed with spectral grade KBr at a ratio of 1:50. The mixture was carefully packed into a macro-cup sampling accessory, and the spectra were recorded from 4,000 to 650 cm À1 at a 4 cm À1 resolution. For each sample, we collected 40 interferograms, which were averaged and corrected against the background spectrum of pure KBr. The spectra were baseline corrected and transformed with the Kubelka-Munk function using ACD Spec Manager (Advanced Chemistry Development, Ontario, Canada). Spectral assignment was performed based on pure standards and according to Silverstein et al. (2005), and peak interpretation was based on Lammers (2008), Lammers et al. (2009), and Movasaghi et al. (2008). We selected 17 identifiable peaks that corresponded to major functional groups for statistical analysis. We calculated the relative peak heights of the functional groups in each sample for comparison across samples from different climate treatments. The relative peak heights were computed as the ratio of the intensity of each peak to the sum of the intensities of all selected peaks (Haberhauer and Gerzabek 1999). We conducted a principal component analysis (PCA) of the relative peak heights to interpret the DRIFT spectra (Suseela et al. 2013).

Data analysis
The different extractable polar metabolite groups such as amino acids, phenolic acids, organic acids, sugars and sugar alcohols were grouped by adding the concentration of individual compounds that were normalized with ribitol. The data were range scaled for further analyses. The metabolomics data of each species were first analyzed using multivariate analysis of variance (MANOVA) which permitted the analyses of all variables in a single statistical method and the evaluation of main and interactive effects of treatments within the experimental design (Johnson et al. 2007, Rivas-Ubach et al. 2012). The significant effects (P , 0.05) from MANOVA were further analyzed using univariate analysis of variance (ANOVA). The univariate analyses were used to identify the metabolite groups that caused differences among treatments (Rivas-Ubach et al. 2012). The main and interactive effects of warming and CO 2 on ester-bound phenolics were analyzed using PROC MIXED. Extractable polar metabolite data, DRIFT spectroscopy and ester-bound phenolics data were analyzed via principal component analysis (PCA; Tharayil et al. 2011). In all statistical analyses differences among individual treatments was determined using Tukey's HSD multicomparison test. All analyses were done using SAS (SAS version 9.2, SAS Institute, Cary, North Carolina, USA).

Extractable polar metabolites
Multivariate analysis of variance showed significant effects of treatments on the different v www.esajournals.org groups of extractable polar metabolites in B. gracilis exposed to different warming and CO 2 treatments (P , 0.001). The important metabolite groups such as phenolic acids, sugar alcohols, sugars, and amino acids responded differently to warming and elevated CO 2 (univariate ANOVA; Fig. 1A-D) while organic acids marginally differed between the warming treatments (P ¼ 0.06; Fig. 1E). Warming increased the relative amounts of phenolic acids (P , 0.001; Fig. 1A) and sugars (P ¼ 0.025; Fig.  1B) but decreased the contents of sugar alcohols v www.esajournals.org (P ¼ 0.004; Fig. 1C). The B. gracilis exposed to the elevated CO 2 treatment showed a higher relative phenolic acid content (P ¼ 0.032; Fig.  1A) and a lower sugar content ( Fig. 1B; P ¼ 0.046). The relative amount of amino acids responded to the interactive effect between the warming and CO 2 treatments, where plants exposed to warming plus ambient CO 2 treatment exhibited the highest relative content of amino acids (P ¼ 0.024; Fig. 1D). These results were further supported by the PCA of extractable metabolite groups (Fig. 2). The PC 1 axis, which explained 44% of the variance in the data, statistically separated warming treatments from the unwarmed treatments. B. gracilis exposed to warming with ambient CO 2 had a higher relative abundance of sugars and those exposed to warming with elevated CO 2 had a higher abundance of phenolic acids. However, the abundance of sugar alcohols and organic acids were found to be associated with the unwarmed treatments. Along PC 2 axis, which explained 24% of the observed variance, B. gracilis exposed to elevated CO 2 treatments were separated from ambient CO 2 treatments. The elevated CO 2 treatments had a higher relative abundance of phenolic acids.
In P. smithii, MANOVA showed significant effect of treatments on the different metabolite groups (P ¼ 0.01). However, only sugars and sugar alcohols varied with different treatments where the levels of sugar alcohols decreased with warming (univariate ANOVA; P ¼ 0.039; Fig. 3B), while those of sugars increased with warming (P ¼ 0.002) and decreased under elevated CO 2 (P ¼ 0.009; Fig. 3C). Organic acids, phenolic acids and amino acids did not differ between the treatments (P . 0.05; Fig. 3). The PCA of the Fig. 2. Principal component analysis of the relative intensities of the extractable polar metabolite groups (amino acids, organic acids, phenolic acids, sugars and sugar alcohols) of B. gracilis subjected to different climatic treatments (ct, ambient CO 2 þambient temperature; cT, ambient CO 2 þwarming; Ct, elevated CO 2 þambient temperature; CT, elevated CO 2 þelevated temperature). The first and second PC axes explained 68.4 % of the variance in the data. Significant treatment effects (P values from ANOVA of principal component analysis) for PC1 and PC2 axes are shown as inset. Letters 'A' and 'B' indicate Tukey's difference between treatments separated by PC1 axis and 'a' and 'b' between treatments separated by PC2 axis.
v www.esajournals.org extractable polar metabolite groups of P. smithii that explained 61% of the variance in the data revealed that warming with ambient CO 2 increased the relative abundance of sugars separating this treatment from the control (ct) treatment that had a higher relative abundance of sugar alcohols (Fig. 4).

Ester-bound phenolic acids
Out of more than 20 phenolic compounds detected, ferulic acid and p-coumaric acid represented .70% of the total phenolic compounds in both species (Fig. 5). The total phenolic acid content in B. gracilis increased with the main Fig. 3. Effect of warming and CO 2 treatments on the relative contents of (A) phenolic acids, (B) sugars, (C) sugar alcohols, (D) amino acids and (E) organic acids in P. smithii. The presented values represent the means 6 SE of the treatment replicates (n ¼ 5). Letters 'a' and 'b' indicate Tukey's difference between warming treatments and letters 'A' and 'B' between CO 2 treatments.
v www.esajournals.org effect of warming (11%; P ¼ 0.001; Fig. 5A) and with the main effect of elevated CO 2 (P ¼ 0.009), while in P. smithii, total phenolic acid content decreased with warming (5%; P ¼ 0.022; Fig. 5B) and increased with elevated CO 2 treatment (7%; Fig. 5B). In B. gracilis, compared to the control, the concentration of ferulic acid increased by 14% with warming under ambient CO 2 (P ¼ 0.011; Fig. 5C), while in the senesced tissues of P. smithii, the concentration of ferulic acid increased with the main effect of elevated CO 2 treatments (8%; P ¼ 0.038; Fig. 5D). Warming treatments did not affect the concentration of ferulic acid in P. smithii.
In B. gracilis, the concentration of p-coumaric acid was highest in the warmed plots with elevated CO 2 (CT) compared to treatments involving warming under ambient CO 2 (cT; 46%; P ¼ 0.004), elevated CO 2 and ambient temperature (Ct; 27%; P , 0.0001), or ambient CO 2 and ambient temperature (ct; 35%; P ¼ 0.002; Fig. 5E). However, in P. smithii, the main effect of warming decreased the concentration of p-coumaric acid by 16% relative to plants in the ambient temperature treatments (P , 0.001; Fig.  5F) and elevated CO 2 increased the concentration of p-coumaric acid (5%; P ¼ 0.026; Fig. 5F) compared to ambient CO 2 treatments.
The results of the principal component analysis of the relative concentrations of ester-bound phenolic acids varied between the species. For B. gracilis, the PC 1 axis, which explained 53% of the observed variance, statistically separated plants exposed to warming with elevated CO 2 from all other treatments (P , 0.001; Fig. 6A) and was associated with a higher relative abundance of 4-hydroxybenzoic acid, syringic acid and pcoumaric acid. Along PC axis 2, which explained 17% of the variance in the data, the B. gracilis in the warmed plots under ambient CO 2 were significantly separated from all other treatments (P ¼ 0.014) and showed a higher relative abundance of ferulic acid, m-coumaric acid and vanillin. Fig. 4. Principal component analysis of the relative intensities of the extractable polar metabolite groups (amino acids, organic acids, phenolic acids, sugars and sugar alcohols) of P. smithii subjected to different climatic treatments (ct, ambient CO 2 þambient temperature; cT, ambient CO 2 þwarming; Ct, elevated CO 2 þambient temperature; CT, elevated CO 2 þelevated temperature). The first and second PC axes explained 57.1 % of the variance in the data. Significant treatment effects (P values from ANOVA of principal component analysis) for PC1 and PC2 axes are shown as inset. Letters 'A' and 'B' indicate Tukey's difference between treatments separated by PC1 axis.
v www.esajournals.org For P. smithii, PC axes 1 and 2 explained 35 and 20% of the variance in the data, respectively (Fig.  6B). The PC 1 axis statistically separated plants in warmed plots exposed to both ambient (cT) and elevated CO 2 (CT) from the ambient temperature treatments (ct, Ct; P , 0.001). The warmed treatments were associated with relatively higher levels of sinapic acid and syringic acid, and the plants in the ambient temperature treatments were characterized by an abundance of acetovanillin and p-coumaric acid.

DRIFT analyses
The initial DRIFT analysis of non-extracted litter of B. gracilis and P. smithii did not reveal any treatment effects; however the DRIFT analysis of  v www.esajournals.org 13 September 2014 v Volume 5(9) v Article 106 the solvent-extracted litter (residual litter after metabolomics analysis) showed significant treatment effects. DRIFT analysis targets vibrational motions in covalent bonds, and along with the cell wall components, a large portion of the extractable phenolics, other aromatic compounds and carbohydrates present in the plant tissue are expected to contribute to these molecular vibrations. Thus, removing the extractable metabolites from the litter matrix helped us to identify the effect of climate on the compositional differences of the non-extractable structural matrix of the litter.
Principal component analysis of the relative intensities of the DRIFT peaks in the residual litter obtained after solvent extraction revealed the effect of warming and CO 2 on the nonextractable structural components of the litter (Fig. 7). For B. gracilis, PC axes 1 and 2 explained 62 and 21% of the variance in the data, respectively (Fig. 7A). Along PC axis 1, the B. gracilis exposed to warming with ambient CO 2 (cT) differed significantly from those in the ambient temperature treatments exposed to either ambient CO 2 (ct) or elevated CO 2 (Ct; P , 0.005). The B. gracilis exposed to warming together with elevated CO 2 also differed significantly from those exposed to elevated CO 2 alone. Warming increased the relative abundance of alkyl compounds from waxes and cuticularmatrix (2851 cm À1 [CH 2 asymmetric stretch], 1424 cm À1 [C-H deformation]) and lignin (1515 cm À1 [aromatic C-C stretch]) and decreased the abundance of carbohydrates (1106, 1059, 1160 cm À1 [combination of C-O stretching and O-H deformation]).
For P. smithii, PC axes 1 and 2 explained 62 and 15% of the variance in the data, respectively (Fig.  7B). The combined effects of warming and elevated CO 2 increased the relative abundance of alkyl carbon (2918 cm À1 , 1424 cm À1 , 2897 cm À1 ) and lignin (1515 cm À1 , P ¼ 0.003) compared with the other treatments. The P. smithii that experienced elevated CO 2 without warming (Ct) were characterized by a higher relative abundance of carbohydrates (1059, 1106, 1160 cm À1 ).

DISCUSSION
In response to environmental stress, plants modify the pathways that regulate the biosyn-thesis and resorption of metabolites, resulting in an altered chemical composition of senesced tissues. Given that grasslands are important sinks for soil carbon and that the majority of them contain a mixture of C3 and C4 species, obtaining a better understanding of the changes in litter chemistry associated with these functional types due to warming and elevated CO 2 is critical to predict carbon storage in grasslands under future climatic conditions. Here, we report that P. smithii (C3) and B. gracilis (C4) in a semiarid grassland exhibit differential litter chemistry when exposed to a factorial combination of warming and elevated CO 2 potentially due to warming induced reduction in soil moisture. The response of extractable compounds did not exhibit a clear relationship with climate treatments, possibly due to the concomitant effect of climate on production and resorption of these mobile metabolites. In B. gracilis the total cell-wall bound phenolic acids increased with the main effect of warming, ferulic acid increased with cT treatment and p-coumaric acid increased with CT treatment and structural compounds such as lignins, waxes and cuticular-matrix showed greater abundance with warming in the absence of elevated CO 2 . Thus in general, warming treatments elicited greater physiological response in C4 species. Conversely, P. smithii (C3) in warmed treatments had lower total cell-wall bound phenolic acids and pcoumaric acid, and only warming accompanied by an increase in CO 2 resulted in an increased concentration of structural components such as lignins, waxes and cuticular-matrix, indicating a better physiological response of P. smithii to moderately stressful climates.

Response of extractable polar metabolites to warming and elevated CO 2
In B. gracilis the relative abundance of extractable polar metabolites such as sugars and free phenolic acids increased with the main effect of warming, and amino acids increased under warming with ambient CO 2 (Fig. 1). High temperature and the accompanying moisture stress (25% reduction in soil volumetric water content compared to elevated CO 2 ; Morgan et al. 2011) often translate into osmotic stress in plants (Szabados et al. 2011), and can induce several types of physiological and biochemical changes, resulting in the altered production of metabolites (Ahuja et al. 2010). Plants subjected to heat and drought stress have been shown to accumulate sugars (Guy et al. 2008) that act as osmoregulants (Rivas-Ubach et al. 2012) and phenolic compounds that function as antioxidants (Urquiaga and Leighton 2000). We interpret the observed changes in extractable polar metabolites in B. gracilis as an adaptive response, as previous investigations performed at the same site as the present study have shown 23% higher biomass production of this species under warming treatments (Morgan et al. 2011).
In P. smithii, only sugars showed a greater relative abundance in response to the main effect of warming (Fig. 3). Thus, the response of extractable metabolites in both the species did not confirm our initial hypothesis. The relative change in abundance of different polar metabolites in the litter of B. gracilis and P. smithii observed in the present study ( Figs. 1 and 2) could be a function of the climatic effect on both production and resorption. Also soil moisture availability, a critical factor determining the resorption of nutrients during senescence, (del Arco et al. 1991) would potentially influence the observed metabolite profiles in the senesced grass tissues.
Response of ester-bound phenolics and structural components to warming and elevated CO 2 Among the total phenolic compounds in plants, those associated with cell wall (esterbound compounds) exhibit a better protective function than those stored in vacuoles (Antonelli et al. 1998, Fischbach et al. 1999 and are not affected by resorption. Ester-bound hydroxycinnamates and their dimers provide structural stability in the cell walls of grasses (Harris and Trethewey 2010), act as initiation sites for lignin and cross-link polysaccharides to lignin (Carpita 1996, Hatfield et al. 1999. In general, in B. gracilis (C4), we observed an increase in the total esterbound phenolic contents, ferulic acid and pcoumaric acid in the warmed treatments ( Fig.  5A, C, E). These cell wall-bound phenolics have been found to increase in plants exposed to drought, thus increasing resistance to moisture stress (Hura et al. 2011). In addition, by acting as a physical barrier against fungal penetration, cell wall-bound phenolics confer improved disease resistance (de Ascensao andDubery 2003, Santiago et al. 2009). Similarly, due to their high UV absorption capacity, cell wall-bound phenolics can attenuate UV-induced oxidative damage to the photosynthetic apparatus (Landry et al. 1995, Schweiger et al. 1996, Hura et al. 2009), to which plants are more prone under drought (Garcia-Plazaola and Becerril 2000). Thus, climatic stress may have led to the greater production of cell wall-bound phenolics in B. gracilis, which in turn, could protect this species against water stress, irradiation and pest and pathogen attack. Thus, B. gracilis exhibited consistency in responding to warming treatments, and was abundant in both structural (Fig. 5A, C, E) and non-structural ( Fig.  1) metabolites and also exhibited a 23% higher biomass production under warming (Morgan et al. 2011). However, in P. smithii the total cell wallbound phenolics, ferulic and p-coumaric acids only increased with the main effect of CO 2 and total cell wall-bound phenolics and p-coumaric acid decreased with the main effect of warming (Fig. 5B, D, F), indicating a metabolic response of this C3 species only under elevated CO 2 .
The DRIFT spectra obtained for B. gracilis and P. smithii captured the differential responses of structural compounds in litter of these two species to the climatic treatments. The B. gracilis plants exposed to warming exhibited a higher relative abundance of lignin, waxes and cuticular-matrix compared to plants exposed to the ambient temperature treatment, which showed a greater relative abundance of carbohydrates (Fig.  7A). Additionally, in combination with elevated CO 2 , the warming treatment produced a marginally higher relative abundance of lignin, waxes and cutin in B. gracilis. Warming in conjunction with elevated CO 2 is thought to pose a greater stress for C4 plants due to the high leaf temperatures caused by the partial closure of stomata . Along with cutin, which forms a water-impermeable layer on leaf surfaces, increased lignin contents, especially the stress lignin formed from acylated monolignols (del Rio et al. 2007), could represent an adaptation to moisture stress (Bargel et al. 2006, Lee et al. 2007, Kosma et al. 2009).
In P. smithii, a similar relative abundance of lignin and cuticular materials was observed only in plants exposed to warming in conjunction with elevated CO 2 (Fig. 7B). Because P. smithii is a cool-season grass, the greater heat/drought stress experienced by this C3 grass when subjected to warming alone might have been overly disruptive to physiology and could have suppressed any metabolic response. This scenario is supported by the observation that in C3 species the metabolic inhibition occurs when the relative water content in the leaves falls below 70%, and CO 2 assimilation rates fail to recover even after removal of water stress (Lawlor 2002, Flexas et al. 2004. Previous studies from the same site have reported similar response of biomass in C3 species to climate treatments, where their biomass production increased by 34% only in elevated CO 2 treatments, but did not respond to warming (Morgan et al. 2011). Thus, the increase in relative abundance of structural v www.esajournals.org metabolites in P. smithii could be construed as a response that occurs only when warming is supplemented with elevated CO 2 , and could partially be attributed to low moisture stress in these treatments (25% higher soil volumetric water content than warming alone treatments; Morgan et al. 2011).
Overall, under a warmer climate, B. gracilis (C4) responded by allocating more carbon for the production of non-structural osmoregulatory compounds, bound phenolics and protective structural components, including lignins, waxes and cutin, whereas a similar response in structural components (lignins, waxes and cutin) in P. smithii (C3) was initiated only when warming was supplemented with elevated CO 2 . Because the described responses of C3 and C4 species clearly coincide with higher biomass production in these species in the respective climates at this study site (Read andMorgan 1996, Morgan et al. 2011), these changes in structural and nonstructural components could represent a potential adaptive response to the respective climates. Thus, our observations regarding the variation in the chemical composition of leaf litter provide a novel, finer, metabolite-level, explanation for a widely known plant-level response that is traditionally quantified by biomass production.

Implications of metabolic changes on the decomposability of tissues
The organismal-level responses to environmental stresses have generally been elucidated by profiling extractable polar metabolites in previous studies (Du et al. 2011, Riikonen et al. 2012, Yu et al. 2012. However, only ,10% of the compounds in plant litter could be extracted by solvents. A major proportion of carbon (;50% of dry mass) resides in the structural cellular components (Vogel 2008), which also undergo quantitative and qualitative changes in response to biotic and abiotic cues. Thus, the present study, which examined both structural and nonstructural components of tissues obtained after sequential extraction, provides an unprecedented, comprehensive understanding of the overall metabolite-level response of plants to climatic changes.
Our analyses of P. smithii and B. gracillis exposed to warming and CO 2 treatments revealed the climatic stress-induced changes in plant structural and non-structural metabolites that could have significant implications for the digestibility and decomposability of these tissues. The higher concentration of extractable polar metabolites observed under warming in B. gracillis and P. smithii could make these tissues more palatable. However, the abundance of ester-bound phenolic compounds, and the structural components such as lignins, cutins and waxes could affect the overall nutritive quality, as phenolic compounds are toxic to microbes in the rumens of animals (Akin 1982). In grasses, the major phenolic acids such as ferulic acid and pcoumaric acid are ester and/ or ether linked to lignin and these cell wall bound phenolics also cross-link polysaccharides to lignin that provides structural integrity to cell walls (Carpita 1996, Hatfield et al. 1999) making the litter more resistant to decay (Grabber et al. 2004, Suseela et al. 2014). Hence, the higher relative abundance of the cell wall-bound phenolics may also decrease the rate of litter decomposition. Thus, even though physiological adaptation might enable B.gracillis and P. smithii to produce higher biomass under future climates, because of the metabolite level changes brought about by the physiological responses, the overall nutritive quality and decomposition susceptibilities of this biomass could be significantly different. Although current models predict higher productivity in grasslands under elevated CO 2 alone, and combined warming and elevated CO 2 scenarios (Parton et al., 2007), our results suggests that in the long-term, concurrent increase in plant structural compounds and cell-wall bound phenolics would reduce the rate of litter decomposition and nutrient cycling leading to a negative feedback to productivity. To our knowledge, this is the first study that captures the differential response of plant metabolites in the senesced tissues of C3 and C4 exposed to factorial combinations of warming and elevated CO 2 .
Although our study reveals the change in litter quality due to warming and CO 2 enrichment, the direct impact of these changes on soil carbon cycling in these grassland ecosystems could be influenced by complex organismal interactions. For example, although litter chemistry is a major driver of soil carbon cycling in ecosystems in the intervening period (Tamura and Tharayil 2014), the soil microbial community could adapt to any v www.esajournals.org climate induced changes in litter chemistry over the long-term through compositional shift. Thus, higher inputs of relatively recalcitrant litter could shift the microbial community to more fungal dominated than a bacterial dominated system that could further alter soil carbon cycling (McGuire andTreseder 2010, Suseela et al. 2013). Changes in litter chemistry at the molecular level may also influence carbon accrual (Tamura and Tharayil 2014) in different SOM pools, as observed in an Australian grassland ecosystem exposed to warming and elevated CO 2 , where higher soil organic carbon and particulate organic carbon contents were observed under C4 species exposed to warming, but not under C3 vegetation .
Our study suggests that if the unique metabolic responses that we observed in B. graciis (C4) and P. smithii (C3), occur in other species of similar functional types, this could also potentially alter the carbon cycling and species distributions in grassland ecosystems.