Elevated field atmospheric CO₂ concentrations affect the characteristics of winter wheat (cv. Bologna) grains

Additional keywords: climate change, candidate core proteins, proteomics.

Over the last 200 years, human activities have accelerated global changes in climate. One of the most relevant of these changes is the increase in atmospheric concentration of carbon dioxide, which, since 1750, has risen from ~280 to 400 ppm, and it is expected to reach ~550 ppm by mid-21st Century (IPCC 2013). This increase is expected to have a great impact on the growth and development of crops, particularly C₃ plants such as wheat. DaMatta et al. (2010) reviewed the literature concerning the effects of CO₂ increase on crop physiology, suggesting some beneficial effects such as yield increase and better water-use efficiency. On the other hand, these beneficial effects can be offset by a decrease in grain quality (Nuttall et al. 2017). In addition, when moving from greenhouse and open-top chamber experiments to the more realistic situation of the free-air CO₂ enrichment (FACE) experiment, the interaction among elevated CO₂ and other environmental variables was shown to introduce higher levels of complexity in plant response (DaMatta et al. 2010). Many studies have focused on the effects of elevated concentrations of CO₂ on the technological and nutritional quality of wheat grains, showing that elevated CO₂ can alter C and nitrogen (N) metabolism in plants, thereby altering the chemical composition of vegetative plant parts (Cotrufo et al. 1998; Loladze 2002). These conditions affect the redistribution and availability of metabolites for the developing grain, subsequently affecting both grain yield and quality (Kimball et al. 2001; Fernando et al. 2014). Some FACE experiments demonstrated only a slight decrease in grain protein levels of bread wheat under elevated CO₂, whereas others pointed to consistent and statistically significant reductions in protein and micronutrient contents (Kimball et al. 2001; Wieser et al. 2008; Högy et al. 2009a; Erbs et al. 2010; Fernando et al. 2012a, 2012b, 2014; Tausz et al. 2013; Myers et al. 2014; Nuttall et al. 2017). In durum wheat, elevated CO₂ decreased the quality of grains and derived pasta, the main negative effect being on gluten content (Fares et al. 2016). Nuttall et al. (2017) reviewed the current knowledge of effects of high temperature and elevated environmental CO₂ on wheat-grain functional properties, and concluded that elevated CO₂ concentrations generally lead to a reduction in protein percentage and a shift in grain composition, with consequent negative impact on bread-making characteristics. However, those authors emphasise that differing results have been obtained among FACE experiments because of the interaction between CO₂ concentration and other environmental factors. CO₂ enrichment was shown to affect even the proteome profile of wheat grains under well-watered, temperate conditions (Högy et al. 2009b) and Mediterranean rainfed conditions (Fernando et al. 2015) in FACE facilities during a single growing season.

The aim of the present work was to investigate effects of elevated CO₂ on yield and grain quality of bread wheat cultivated in a Southern European environment. The winter wheat (Triticum aestivum L.) genotype Bologna has been used as model because of the prevalence of winter cereal cultivation in several warm temperate and Mediterranean areas. Bologna is representative of frost-resistant winter wheats well suited to exploit long growing seasons with autumn sowing, not only for cool winters but also in warm continental and mountain climates subject to random frost episodes in the context of the warming climate, as in the Po Valley. Using a FACE facility, we exposed wheat plants to 570 ppm CO₂, which simulates the terrestrial tropospheric concentration predicted for 2050 (www.ipcc.ch/). This treatment, e[CO₂], was compared with an ambient control, a[CO₂]. The experiments were performed for two growing seasons to consider effects under the environmental conditions encountered during different years. Key characteristics related to grain quality were considered, including contents of N, free amino acids and minerals, and composition of seed-storage proteins. Seed proteome profiles were obtained to look for proteins that were differentially expressed between a[CO₂] and e[CO₂] in both seasons, and which therefore represent good candidates to be directly affected by elevated CO₂ in wheat (core proteins) independent of interactions of CO₂ with prevailing environmental conditions.

The superior bread-making wheat cv. Bologna was used. Bologna is currently the most cultivated variety in Italy owing to its high yield stability and excellent quality. It has a winter growth habit, characterised by a wide range of adaptability, high levels of cold and lodging resistance, and high tolerance to fusarium head blight (FHB), and is classified as ‘Frumento di Forza’ (FF, Improver Wheat) from a technological point of view. Cultivar Bologna (pedigree H89092/H89136//Soissons) has been bred by SIS, Società Italiana Sementi, San Lazzaro di Savena, BO, Italy (www.sisonweb.com/it/index.php).

Wheat plants were grown within the FACE facility of the Research Centre for Genomics and Bioinformatics (CREA-GB) at Fiorenzuola d’Arda (44.927°N, 9.893°E), in the Po Valley (Italy), at elevation 60–80 m a.m.s.l. and with a warm continental climate. The field soil is alkaline (pH 8.09), with the following characteristics: total carbonate, 10.19%; inorganic C, 12.22 g kg^–1; total C, 28.1 g kg^–1; organic C, 15.9 g kg^–1; organic matter, 2.74%; total N, 0.10%; C : N ratio 15.6; P₂O₅, 21.7 mg kg^–1; cation exchange capacity, 6.85 cmol(+) kg^–1.

Experimental units were plots 2.2 m by 1.36 m. The FACE treatment, with the e[CO₂] target value set at 570 ppm, was replicated in four octagons inscribed in circles of 14 m diameter. The a[CO₂] controls were replicated four times in octagons without FACE. Within the octagons, plots were replicated four times in the first growing season (2011–12; total 4 × 4 = 16 plots per treatment), and two times in the second growing season (2012–13; total 4 × 2 = 8 plots per treatment). Designations Y1E, Y1A, Y2E and Y2A refer to year (Y1, 2011–12; Y2, 2012–13) × CO₂ treatment (E, e[CO₂]; A, a[CO₂]) combinations.

Apart from the CO₂ fumigation, plants were treated according to local agronomic practices. In 2011–12, the sowing, start of fumigation and harvest dates were 19 October 2011, 16 November 2011 and 2 July 2012, respectively. FACE treatment was stopped when leaves were senescent on 14 June 2012. FACE treatment was interrupted for 20 days in February 2012 when the plots were covered with snow. Plots were fertilised with application of an N-P-K fertiliser at pre-seeding and two topdressings with ammonium nitrate for a total of 149 kg N ha^–1.

In 2012–13, the sowing, start of fumigation and harvest dates were 24 October 2012, 9 November 2012 and 11 July 2013, respectively. Plots were again fertilised with application of N-P-K fertiliser at pre seeding and two topdressings with ammonium nitrate. However, the fertiliser doses were increased to a total of 234 kg N ha^–1 because, in the first season, the export of N with harvested grains and straw had exceeded the amount applied as fertiliser, and the experiment was planned as non-nutrient-limited.

Air temperature, precipitation, relative humidity, and global radiation were measured and recorded at 10-min intervals with an automatic meteorological station located within the field site of the FACE experiment at Fiorenzuola.

After harvest, aboveground biomass, grain yield, 1000-grain weight and harvest index were determined. Grain N content was determined by the Kjeldahl method, and grain crude protein content calculated as 5.7 × N. Grain CHN content was determined by TruSpec CHN elemental analyzer (LECO Corp., St. Joseph, MI, USA) following the manufacturer’s specifications. Grain N yield (kg ha^–1) was calculated from grain yield and grain N content. These traits were determined for every plot.

For each growing season, samples for each treatment were obtained for further analyses from the pool of grains harvested in four a[CO₂] and four e[CO₂] experimental plots. For each analysis, at least three technical replicates were performed.

Wheat kernels were ground by mortar and pestle, using liquid nitrogen. To avoid metallic contaminations, a metallic grinder was not used. A pulverised sample (~200 mg) was accurately weighted inside a pressure-resistant vessel (100 mL) and digested with 3 mL 69% concentrated, ultrapure nitric acid and 9 mL 36% concentrated, ultrapure hydrochloric acid. The reactor was closed, starting the digestion procedure in a microwave oven following the program: step 1, 250 W for 2 min; step 2, 0 W for 1 min; step 3, 250 W for 2 min; step 4, 400 W for 2 min; step 5, 500 W for 5 min; step 6, 0 W for 1 min; step 7, 600 W for 5 min; step 8, oven ventilation for 6 min. After cooling the vessels to room temperature, each sample was diluted to 20 mL in a low-density polyethylene container with ultrapure water produced by a Millipore Milli-Q Advantage A10 system (Merck, Darmstadt, Germany). Ultrapure nitric acid (600 µL 69% concentrated) was added to the solution to obtain a final concentration of 2% nitric acid. The final solution obtained was analysed by inductively coupled plasma-optical emission spectroscopy (iCAP 6300; ThermoFisher Scientific, Waltham, MA, USA) for the determination of Ca, Cu, Fe, K, Mg, Mn and Zn. Determination of trace elements was performed by using external calibration.

Wheat flour was extracted with 80% cold ethanol in the presence of nor-Leu (50 nmol) as an internal standard, homogenised with a Teflon pestle, and centrifuged at ~14 000g at 4°C. The supernatant was lyophilised, treated with 3% sulfosalicylic acid (500 µL) to precipitate any protein fraction still present, and centrifuged again (Iriti et al. 2009). Then 30 µL of this extract was directly analysed. Each sample was individually prepared and analysed in triplicate.

Aliquots of samples were directly analysed on a Biochrom 30 amino acid analyser (Biochrom, Cambridge, UK) equipped with a post-column ninhydrin derivatisation system, adapting the procedure of Moore and Stein (1963). Total free amino-acid content was obtained with a single analysis, simplifying the acquisition of raw data.

For the free amino-acid profiles, standard codes were used, along with other abbreviations, as follows: AAAA, L-α-aminoadipic acid; Ala, L-alanine; Arg, L-arginine; Asn, L-asparagine; Asp, L-aspartic acid; Ethan, ethanolamine; GABA, γ-amino-butyric acid; Gln, L-glutamine; Glu, L-glutamic acid; Gly, glycine; His, L-histidine; Ile, L-isoleucine; Leu, L-leucine; Lys, L-lysine; Met, L-methionine; Orn, L-ornithine; Pea, o-phosphoethanolamine; Phe, L-phenylalanine; Phser, o-phospho-L-serine; Pro, L-proline; Sarc, L-sarcosine; Ser, L-serine; Taur, taurine; Thr, L-threonine; Trp, L-tryptophan; Tyr, L-tyrosine; Val, L-valine; β-ala, β-alanine.

Seed protein was extracted according to Wang et al. (2006) with minor modification. Briefly, 1 g finely powdered seeds was suspended in 10 mL iced 10% w/v TCA–acetone, vortexed, filtered with Miracloth, and centrifuged at 16 000g for 10 min at 4°C. After washing once with iced 80% 0.1 m ammonium acetate in methanol and twice with 10 mL iced 80% acetone, pellets were resuspended in 5 mL ice-cold sodium dodecyl sulfate (SDS) extraction buffer (30% w/v sucrose, 0.1 m Tris-HCl pH 8, 2% w/v SDS, 2% v/v β-mercaptoethanol, 1.5% w/v polyvinylpyrrolidone, 1 mm PMSF, 1 mm cocktail antiprotease) and then extracted with an equal volume of phenol-saturated 500 mm Tris-HCl, pH 8. Proteins were precipitated overnight by adding five volumes cold 0.1 m ammonium acetate in methanol, and solubilised in rehydration sample buffer (7 m urea, 2 m thiourea, 4% (w/v) CHAPS detergent, 1% (v/v) Triton X-100, 20 mm Tris, 1% (w/v) DTT (dithiothreitol), 0.2% (w/v) ampholine 4-7). Protein concentration was estimated according to the Bradford (1976) assay. Three independent protein extractions were performed from each sample.

Total proteins (500 mg) were loaded onto an 18-cm, pH 4–7 linear gradient or pH 3–10 nonlinear gradient (NL) Immobiline DryStrip gels (IPG strips) (GE Healthcare, Uppsala, Sweden), and separated as described by Marsoni et al. (2008). The SDS-PAGE gels were visualised by the modified Colloidal Coomassie Brilliant Blue staining method (Aina et al. 2007). Each separation was repeated three times for each biological replicate to ensure the protein pattern reproducibility.

Gels were digitalised using GS-800 imaging systems (Bio-Rad, Munich, Germany), and the images (TIFF format, 300 dpi, 16 bit) were analysed with PDQuest software version 8.0.1 (Bio-Rad). Only the protein spots present in all gels for each sample were considered for the statistical analysis. Statistical analysis (Student’s t-test at P = 0.05) identified proteins that significantly increased or decreased (≥1.5-fold in relative abundance) after the different treatments with respect to the control. These spots were selected for tandem mass spectrometry (MS/MS) analysis.

Selected spots were manually excised from the 2D gels, washed twice and stored in 50% ethanol at 4°C until digestion. Spot digestion was performed as previously described (Marsoni et al. 2008). The extracted tryptic fragments were analysed by MS/MS after reverse-phase separation of peptides by liquid chromatography–electrospray ionisation MS/MS (LC-ESI-MS/MS). All experiments used a Finningan LXQ linear ion-trap mass spectrometer, equipped with a Finningan Surveyor MS plus HPLC system (Thermo Electron Corp., Waltham, MA, USA).

Chromatography separations were conducted on a BioBasic C18 column (150 mm internal diameter, 150 mm length, 5 mm particle size; Thermo Electron Corp.), using a linear gradient from 5% to 75% acetonitrile, containing 0.1% formic acid, for 50 min with a flow of 2 mL min^–1. Acquisitions were performed in the data-dependent MS/MS scanning mode (full MS scan range of 400–1400 m/z followed by zoom scan for the most intense ion from the MS scan and full MS/MS for the most intense ion from the zoom scan), thus enabling a dynamic exclusion window of 3 min.

Protein identification was performed by searching in the National Center for Biotechnology Information (NCBI) viridiplantae and/or EST-viridiplantae protein database using the MASCOT program (www.matrixscience.com). The following parameters were adopted for database searches: complete carbamidomethylation of cysteines, partial oxidation of methionines, peptide mass tolerance 1.2 Da, fragment mass tolerance 0.8 Da, and missed cleavage 1. For positive identification, the ion score (–10log(P)) had to be over the significance threshold level of P = 0.05. Spectra were also compared with the in situ database by SEQUEST algorithm incorporated in BIOWORKSBROWSER 3.3 software (ThermoFisher Scientific) against the Triticum subset database (31 852 entries) obtained from the NCBI-nr database (https://www.ncbi.nlm.nih.gov/) and, in case of mismatch, against the full NCBI-nr database (25 877 237 entries). The searches were carried out as described in Marsoni et al. (2008). If needed, the name of unknown proteins was annotated by protein similarity search performed by alignment analysis against the NCBI-nr database by using BLAST (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

Non-identified spectra were submitted to de novo analysis by PepNovo software, using default parameters (http://proteomics.ucsd.edu/Software/PepNovo). Only those PepNovo results that received a mean probability score of at least 0.5 were accepted. Peptide candidates obtained by PepNovo were edited according to MS BLAST rules, and an MS BLAST search was performed against the NCBI-nr database at: http://genetics.buh.harvard.edu/msblast. Statistical significance of hits was evaluated according to the MS BLAST scoring scheme. In order to assign a correct identification, we also take into account a minimum of two different matching peptides and correspondence of the molecular weight.

The proteins were extracted as described above. Proteins (20 µg) were derivatised with DNPH (2,4-dinitrophenylhydrazine) as previously described with some modifications (Levine et al. 1994). Briefly, the proteins were denatured adding SDS at the final concentration of 6%. The derivatisation was performed by adding one volume of 10 mm DNPH in 2 n HCl. Only 2 n HCl was added to the negative control. After incubation for 30 min at room temperature, the mixture was neutralised by adding one volume of neutralisation solution (2 m Tris, 30% glycerol). Proteins were separated by 12% SDS-PAGE and transferred to PVP (polyvinylpyrrolidone) membrane (SERVA Electrophoresis, Heidelberg, Germany). The oxidative modified proteins were detected by using anti-DNP (dinitrophenyl-group) antibodies (Sigma-Aldrich, St. Louis, MO, USA) and visualised by a chemiluminescence detection kit (SuperSignal; Pierce Biotechnology, Rockford, IL, USA). To monitor the equal loading of samples, CCBB (colloidal Coomassie Brilliant Blue) was used to stain the proteins in a duplicate gel. Gel and immunoblot images were acquired by using GS-800 imaging systems and analysed using ImageJ software (Bio-Rad Laboratories, Hercules, CA, USA).

Seeds were surface-sterilised with 10% sodium hypochlorite solution for 10 min and then rinsed with distilled water. Hydrated seeds were transferred in Petri dishes (100 mm diameter) containing a Whatman filter in the presence of 5 mL distilled water, pH 7.1. Ten dishes, each containing 10 seeds, were prepared for each test and incubated in the dark at 25°C. After 2, 4 and 7 days, the germinated seeds were counted. Seeds were considered germinated when root elongation was >3 mm. Tests were performed in triplicate.

Results for grain yield and N content were analysed with analysis of variance (ANOVA) as a split-plot design using R statistical software (R Foundation for Statistical Computing, Vienna: www.R-project.org/). The remaining results were presented as mean of the technical replicates ± standard deviation (s.d.). Differences between treatments for the different measured variables were tested by one-way ANOVA, followed by the Bonferroni-Holm’s post-hoc test (Daniel’s XL Toolbox version 6.60: http://xltoolbox.sourceforge.net).

The climatic conditions during both growing seasons were sufficiently moist, with 487 mm precipitation over the crop growth cycle in 2011–12 and 1027 mm in 2012–13. A developing drought risk in March–April 2012 did not materialise due to rainfall in April. In both growing seasons, the ratio potential evapotranspiration : rainfall remained well below 1 (0.66 and 0.30, respectively). No marked temperature extremes were registered (FW Badeck, F Rizza, C Maré, E Mazzucotelli, L Cattivelli, A Zaldei, F Miglietta, unpubl. data), and no frost damage occurred (Badeck and Rizza 2015).

Total wheat biomass and grain yield were higher under e[CO₂] than a[CO₂] in both years of the study (Table 1). The biomass increase was significant in both years. Grain yield increased by 21% in the two years, significantly in 2011–12 and as a trend (P = 0.0919) in 2012–13. No significant CO₂ treatment effect was found for 1000-grain weight and harvest index, which showed slight increase and decrease, respectively. Heading date did not significantly differ between CO₂ treatments in either year. The grain N content (%) was reduced under e[CO₂], showing a non-significant trend in the first year, and a significant effect in the second. However, a positive, although non-significant trend was evident when comparing grain N yield (kg ha^–1), with an increase of 11.9% in Y1E vs Y1A and an increase of 8.6% in Y2E vs Y2A.

Table 1.  Effects of elevated CO₂ concentration (e[CO₂], 570 ppm) vs ambient (a[CO₂]) on biomass and grain properties of winter wheat cv. Bologna during two growing seasons
*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; n.s., non-significant (P > 0.05)

A negative trend was observed in the early germination rates of seeds from plants grown under e[CO₂] compared with the control (Fig. 1).

Fig. 1.  Germination rate of winter wheat cv. Bologna seeds. Y1, 2011–12; Y2, 2012–13; E, elevated CO₂ (570 ppm); A, ambient CO₂. Capped lines are ± standard deviation. *P < 0.05.

No significant differences in free amino-acid contents of individual grains were found between CO₂ treatments in the first year (Table 2). In the second year, the grain content of free Ala, Arg, Asn, Glu, Gln, Leu, Lys, Pro, Thr and Val was lower under e[CO₂] than a[CO₂]. However, overall content of free amino acids was higher in Y2A than Y1A, and particularly the relative amounts of Asn, Arg, Asp, Gln, Glu, and Trp (Table 2), whereas Y2E and Y1E had similar amino acid contents. No significant effect of CO₂ concentration or of growing season was observed on grain macro- and micronutrient contents (Table 3), except for a decrease in Zn concentration in samples grown under e[CO₂] compared with the control in both years.

Table 2.  Free amino-acid profile (mg 100 mg^–1 flour) of seed extracts from wheat grown in a FACE system at ambient (a[CO₂]) and elevated (e[CO₂], 570 ppm) CO₂ concentrations
Common amino acids are indicated by three-letter code; other abbreviations: Phser, o-phospho-L-serine; Taur, taurine; Pea, o-phosphoethanolamine; AAAA, L-α-aminoadipic acid; GABA, γ-amino-butyric acid; Ethan, ethanolamine; Orn, L-ornithine; Sarc, L-sarcosine. Values are means ± s.d. (n = 4). Statistical significance determined according to 1-way ANOVA and Bonferroni–Holm’s post-hoc test (Daniel’s XL Toolbox version 6.60, http://xltoolbox.sourceforge.net). In case of failure of test for equal variance, Student’s t-test was used. Level of significance P = 0.05; n.s., not significant (P > 0.05)

Table 3.  Macro- and micro-nutrient concentrations according to growing season and CO₂ concentration treatment
a[CO₂], Ambient CO₂; e[CO₂], elevated CO₂ (570 ppm). Significant differences occurred only for Zn concentration with respect to CO₂ concentration in both years

We analysed total soluble proteins by 2DE with two pH ranges: 4–7 (Fig. 2a, b) and 3–10 NL (Fig. 2c, d). In the pH 4–7 gels, we detected ~577 spots in Y1A, 566 spots in Y1E, 593 spots in Y2A and 629 spots in Y2E. The similarity of the gels was 82.6% for Y1E vs Y1A and 69.3% for Y2E vs Y2A. In the pH 3–10 NL gels, we detected 791 spots in Y1A, 622 spots in Y1E, 591 spots in Y2A and 534 spots in Y2E. The similarity of the gels was 59% for both growing seasons.

Fig. 2.  Representative 2D gels of the control samples grown in (a, c) the first (2011–12) and (b, d) the second (2012–13) growing season. Only the proteins identified are circled. Protein identities correspond to spot numbers in the Supplementary materials tables S1 and S2.

Overall proteomic analysis revealed 52 differentially expressed protein spots in samples grown under e[CO₂] compared with the control in 2011–12, and 53 in 2012–13. Comparison of Y1E and Y1A samples revealed 37 protein spots that were down-accumulated (71%), whereas 15 spots were up-accumulated (29%). Comparison of Y2E and Y2A samples showed that 22 protein spots were down-accumulated (42%) and 31 up-accumulated (58%). The differentially expressed protein spots common to both years are reported in Table 4; these proteins represent good candidates to be directly affected by increased CO₂ concentration in wheat (core proteins).

Table 4.  Common proteins in both growing seasons
a[CO₂], Ambient CO₂; e[CO₂], elevated CO₂ (570 ppm); Mr, relative molecular mass; pI, isoelectric point; exp., expected

The differentially expressed proteins identified by MS/MS were classified into five groups according to their putative physiological functions: metabolism, protein synthesis and modification, detoxification and defence, storage, and unknown proteins. Approximately 35% and 30% of the differentially expressed protein spots in the 2011–12 and 2012–13 samples are metabolic proteins, and ~35% and 42% are storage proteins, respectively (Supplementary materials tables S1 and S2, available at the journal’s website).

Some protein spots corresponding to high-molecular-weight glutenin subunits (HMW-GS) accumulated during both growing seasons under e[CO₂]. CO₂ enrichment also resulted in the differential expression of several protein spots identified as globulin-3, globulin-3A, globulin-3C, globulin-1 S allele, and globulin-like protein. Each globulin was identified in multiple spots, with some exhibiting a lower relative molecular mass (MW) and more acidic isoelectric point (pI) than the theoretical values. Analysis of sequence coverage indicated that these spots were produced by cleavage of the N-termini of these proteins. These results are consistent with previous investigations indicating that globulin-3 family proteins can be processed at internal cleavage sites to yield polypeptides with a large range of molecular weights and pI values (Singh et al. 2001; Dupont et al. 2011; Koziol et al. 2012). Interestingly, the spots corresponding to full-length globulin were down-accumulated under e[CO₂], whereas those corresponding to endo-proteolytic events accumulated or appeared in samples grown under these conditions.

Proteome analysis revealed that one superoxide dismutase (SOD) accumulated in wheat grains grown under e[CO₂] in both years. Moreover, wheat grains obtained during 2012–13 also accumulated two isoforms of an ascorbate peroxidase (APX). One-dimensional PAGE of seed protein extracts was obtained and the presence of carbonyl groups was detected by western blotting using the DNPH immunoassay. The e[CO₂] treatment resulted in greater accumulation of oxidatively modified polypeptides in wheat grains during both growing seasons, as revealed by an increase in carbonylated protein levels (Fig. 3).

Fig. 3.  Detection of oxidatively modified proteins by: (a) protein stain and (b) anti-DNP immunoassay. (c) Carbonylation index. For (c): Y1, 2011–12; Y2, 2012–13; E, elevated CO₂ (570 ppm); A, ambient CO₂; capped lines are ± standard deviation; *P < 0.05.

Approximately one-third of the protein spots expressed differentially between seeds grown in e[CO₂] and the control are involved in metabolic processes. However, we identified different proteins across the two growing seasons, indicating that the influence of the e[CO₂] treatment on seed metabolic activity is strongly influenced by additional environmental effects (Table 4). Only β-amylase and some isoforms of glyceraldehyde-3-phosphate dehydrogenase were down-accumulated in the Y1E and Y2E samples.

The effects of climate change on our ecosystems are already severe and widespread, and ensuring food security in the face of climate change is among the most daunting challenges facing humankind. Among the actions needed to build resilience to climate change into agricultural production systems is a deeper knowledge of the impact of changed environmental conditions on the yield and quality of cereals. In this context, we have evaluated the agronomic performance and seed quality of an autumn-sown bread wheat variety cultivated in open field conditions in a FACE system, at the CO₂ concentrations expected in 2050. Because interactions between elevated CO₂ and other environmental conditions make it difficult to interpret the impact of increasing CO₂ levels on wheat grains, we compared the responses of wheat during two growing seasons.

In the warm continental climate of the Po Valley, a significant increase in aboveground biomass was observed for cv. Bologna plants grown at elevated CO₂. As reported in Table 1, the biomass increases of 23% and 27% in the first and second year, respectively, are values even higher than reported in similar FACE trials (Högy et al. 2009a). In both years, we observed a 21% yield increase at elevated CO₂. This result is in agreement with the data obtained in several other FACE experiments (DaMatta et al. 2010; Arachchige et al. 2017; Nuttall et al. 2017). An average grain-yield increase of 15% in bread wheat grown at elevated CO₂ in FACE experiments has been found through meta-analyses (Ainsworth and Long 2005). Increases of 10% and of 26% have been reported in German and Australian environments, respectively (Högy et al. 2010; O’Leary et al. 2015).

The first conclusion is therefore a confirmation of the ‘fertiliser’ role of elevated CO₂ even in winter wheat in the Southern European environment considered here.

According to our data, the relative biomass allocation to seeds, the harvest index (HI), was quite stable (Table 1) and only 2% and 4.4% lower under e[CO₂] in 2011–12 and 2012–13, respectively. Variability for this agronomic parameter has been found among durum wheat cultivars grown at different CO₂ levels (Aranjuelo et al. 2013). Those authors observed that different genotypes increase or reduce their growth, and consequently their HI, under elevated CO₂ as a result of different regulation of light-saturated rate of CO₂ assimilation and different N assimilation. Therefore, HI has been identified as a parameter that influences responsiveness to elevated CO₂ in durum wheat (Aranjuelo et al. 2013).

Total protein and N concentrations declined in seeds of cv. Bologna grown under e[CO₂]. This observation is in agreement with other studies demonstrating that the most pronounced impact of elevated CO₂ on nutritional and functional quality of common wheat was the significant reduction in protein concentration by 3.5–14.3% (Kimball et al. 2001; Wieser et al. 2008; Bloom et al. 2010; Fernando et al. 2012a; Broberg et al. 2017). Several hypotheses have been proposed to explain this decline: (i) excess accumulation of carbohydrates leads to reduced grain protein concentration (biomass dilution); (ii) decreased N-uptake rate under elevated CO₂; (iii) elevated CO₂ inhibits N assimilation (Taub and Wang 2008; Bloom et al. 2010). Högy et al. (2009a) hypothesised that the N concentration decrease is a result of the limited N supply from vegetative parts of the plant and that ‘the current rates of N fertilizer are probably inadequate to maintain existing grain quality standards’. However, increases in N fertilisation pose several problems in terms of costs and environmental consequences, and probably result in higher biomass and yield instead of promoting the distribution of N into grains (Högy and Fangmeier 2008). Nowak et al. (2004) agree that N availability has a key role in modulating the plant response to elevated CO₂, even if the physiological and molecular mechanisms of this decrease are not fully understood. Recently, Fernando et al. (2017) tested the hypothesis that N types (NH₄⁺ and NO₃^–) and their ratio can be associated with grain N concentration under elevated CO₂. Those authors found that when equal quantities of NH₄⁺ and NO₃^– are supplied to the plant under elevated CO₂, post-anthesis N uptake is stimulated and the reduction in grain N concentration is avoided.

However, if we take into account the grain N yield, an increase was observed under Y1E and Y2E growth conditions. Högy et al. (2009c) found that the total protein concentration of spring wheat grown at elevated CO₂ was significantly decreased, but that protein yield showed an opposite trend.

Because changes in seed C : N ratio can affect seed vigour (Hampton et al. 2013) and because we have found an increase in C : N ratio in e[CO₂] seeds, germination tests were done, finding that parental CO₂ environment did not have an effect on germination percentage. However, a slower germination was observed in the e[CO₂] seeds. The same effect was observed by Thürig et al. (2003) for seven different graminoids, non-legume forbs and legumes species. Those authors in fact observed that the fraction of germinated seeds was not significantly different between seeds produced at ambient and elevated CO₂ in those species. However, the time to germination of seeds produced under elevated CO₂ was significantly shortened in two species (Sanguisorba minor, Trifolium pratense) and prolonged in Briza media. No difference was observed in the remaining five species. In the annual grass Bromus rubens, the plant population reacted to elevated CO₂ by producing seeds with smaller reserves and lower initial performance (Huxman et al. 1998). Variations in C : N ratio and seed size can be responsible for the difference in seed germination, but other factors may be involved. However, in the present experiment, neither 1000-kernel weight nor grain-size distribution changed significantly under e[CO₂]. The whole-proteome characterisation of cv. Bologna seeds did show the accumulation of oxidatively modified polypeptides under e[CO₂]. The accumulation of APX and other defence-related proteins in spring wheat grains under elevated CO₂ was also reported by Qiu et al. (2008) and Högy et al. (2009b). In plants, SOD and APX are two of the main ROS-scavenging enzymes, and constitute a highly efficient system for removing superoxide and hydrogen peroxide. The observed increase in protective antioxidative mechanisms suggests a higher oxidative level in wheat grains grown under elevated CO₂. ROS can act as damaging or signalling molecules depending on the delicate equilibrium between ROS production and scavenging. If the ROS level exceeds the threshold, cells are exposed to oxidative stress. Several stressful environmental conditions (such as drought, high light levels, pathogens) induce high levels of cellular ROS, which can damage biomolecules such as lipids, proteins and DNA. Protein carbonylation, one of the most harmful, irreversible oxidative modifications of protein, involves the introduction of carbonyl groups into protein-bound amino acids, and it can disrupt protein activity thereby contributing to cell injury. Protein carbonylation is considered a hallmark of oxidative damage; we therefore performed protein carbonyl analysis to assess the extent of oxidative stress under e[CO₂] (Fedorova et al. 2014). The results obtained indicate that pro-oxidant conditions and oxidative stress occur during seed development under elevated CO₂. Seed performance is highly dependent on parental environmental cues, especially during seed formation and maturation (Donohue 2009). Plants of various species exhibit altered antioxidant capacity and an increased oxidative stress under elevated CO₂ (Cheeseman 2006; Qiu et al. 2008; Naudts et al. 2014). This increased oxidative stress under elevated CO₂ might be due to increased canopy temperatures (Kimball et al. 1999; Högy et al. 2009b). Uncontrolled ROS production might have an impact on grain germination (Bykova et al. 2011).

Future work will be necessary to understand the possible significance of elevated-CO₂-induced oxidative stress and the impact of protein carbonylation on grain germination. However, we speculate that the observed decrease in the speed of germination, but not in germination rate per se, could be a consequence of carbonylation of the proteins in addition to reduced seed protein content under elevated CO₂ (Hampton et al. 2013). In summary, parental CO₂ environment seems responsible for variation in seed vigour, a key character for successful crop production and for the future of seed industry.

Moreover, we have compared the quality of the seeds produced in treatments e[CO₂] vs a[CO₂] at the two levels of free individual amino-acid content and storage-protein composition. We observed that the relative amounts of Asn, Arg, Asp, Gln, Glu and Trp were significantly higher only in the Y2A samples. Overall, our data indicate that the differences in N, protein, and free amino-acid contents in e[CO₂] vs a[CO₂] seeds were greater in 2012–13 than in 2011–12. The differences in biomass production and grain yield were consistent between years, suggesting that the observed effect might be due to a lower N uptake or assimilation rate under e[CO₂]. Other studies report a decrease in specific amino acids such as Gln, Lys, Arg and Pro (DaMatta et al. 2010) or a general decrease in both essential and non-essential amino acids (Högy et al. 2010). In any case, the variations of several amino acids highlight a relationship among N translocation compounds–N storage and CO₂ concentration (Arp et al. 1998), considering that Asp, Asn, Glu and Gln are generally regarded as the major N-translocation compounds in plants (Persson et al. 2006) and Arg is often a major N-storage compound in plants (Winter et al. 2015). Noteworthy also is the variation in Trp, a precursor of several tryptophan-derived signals such as indole-3-acetic acid (IAA or auxins), which are involved in most aspects of plant development (Pelagio-Flores et al. 2011). Finally, observed changes in Asn (4C : 2N) vs Gln (5C : 2N) retrieved suggest different efficiency in terms of C usage at different CO₂ concentration (Zheng 2009).

In conclusion, elevated CO₂ and other environmental variables seem to interact, giving a range of minor effects on the wheat grain composition.

The whole proteome of seeds grown under e[CO₂] vs a[CO₂] has also been characterised to obtain better insight into storage protein composition, directly linked to technological properties of wheat seeds. Previous results indicated that elevated CO₂ induces alterations in gluten protein-subunit composition, which, in turn, affects the rheological characteristics of wheat flour. For example, the glutenin : gliadin ratio diminished in winter wheat cv. Batis grown under elevated CO₂ at low N fertilisation and increased at higher N doses (Wieser et al. 2008). Moreover, a 3-year study of three bread-wheat varieties demonstrated a reduction in grain protein content and an increase in the glutenin : gliadin ratio (Panozzo et al. 2014). Our data are in disagreement with previous reports indicating that rising CO₂ levels have a negative impact on HMW-GS content in wheat (Högy et al. 2009b; Fernando et al. 2015); we have observed an accumulation of some HMW-GS under e[CO₂]. These findings suggest that the impact of elevated CO₂ on glutenin subunit accumulation may depend not only on CO₂ level but also on its interaction with the plant genotype or additional environmental variables. Glutenins and gliadins contribute ~80% to grain protein content, with albumin and globulins covering the remaining 20%. In our trials, we have observed a down-accumulation of full-length globulins and up-accumulation of those forms corresponding to endo-proteolytic events. Globulin expression is well known to be environmentally dependent in some wheat cultivars (Altenbach and Kothari 2004; Sancho et al. 2008; Laino et al. 2010; Pompa et al. 2013). This class of proteins is multifunctional, especially in seeds. In addition to their well-established role in nutrient storage, globulins may be important for plant defence against fungi and insects (Macedo et al. 1993; Dunwell et al. 2000; Donaldson et al. 2001; Lane 2002). Various peptide fragments of globulins exhibit antimicrobial activity in several plant species (Chung et al. 1997; Marcus et al.1999; Yamada et al. 1999), and they may play a role in inhibiting the growth of potential pathogens around the germinating seed. The same holds true for other antimicrobial proteins such as a vicilin protein that accumulated in the Y1E samples, and one avenin-like protein and one chitinase that accumulated in Y2E grains. The induction of such proteins cannot be explained simply by the e[CO₂] treatment, but might be linked to the microbial populations that developed in the 2 years. In addition, Arachchige et al. (2017) found a higher abundance of avenin-like seed proteins and of serpin-Z1C like defence proteins in wheat genotypes under elevated CO₂.

In conclusion, the e[CO₂] conditions increased the grain yield and biomass production of a winter wheat (cv. Bologna) grown in a Mediterranean environment and increased its C : N ratio, whereas grain mineral content was not affected, except for a decline in Zn concentration. This Zn decrease has been found in other comparisons involving wheat grains grown at elevated CO₂ (Broberg et al. 2017). Moreover, the e[CO₂] treatment significantly altered the proteome pattern of cv. Bologna grains. However, the low level of overlap of differentially expressed proteins suggests that many of the changes in the grain proteome of the same cultivar are caused by exposure to elevated CO₂ in conjunction with the specific environmental conditions faced during the growing season. The observed changes are therefore highly dependent on interactions with prevailing environmental conditions. This important factor should be taken into consideration when selecting genotypes with desirable resilience to the elevated CO₂ conditions in the future.

Authors declare no conflicts of interest.