Rubisco catalytic properties of wild and domesticated relatives provide scope for improving wheat photosynthesis

Abstract

Rubisco is a major target for improving crop photosynthesis and yield, yet natural diversity in catalytic properties of this enzyme is poorly understood. Rubisco from 25 genotypes of the Triticeae tribe, including wild relatives of bread wheat (Triticum aestivum), were surveyed to identify superior enzymes for improving photosynthesis in this crop. In vitro Rubisco carboxylation velocity (V_c), Michaelis–Menten constants for CO₂ (K_c) and O₂ (K_o) and specificity factor (S_c/o) were measured at 25 and 35 °C. V_c and K_c correlated positively, while V_c and S_c/o were inversely related. Rubisco large subunit genes (rbcL) were sequenced, and predicted corresponding amino acid differences analysed in relation to the corresponding catalytic properties. The effect of replacing native wheat Rubisco with counterparts from closely related species was analysed by modelling the response of photosynthesis to varying CO₂ concentrations. The model predicted that two Rubisco enzymes would increase photosynthetic performance at 25 °C while only one of these also increased photosynthesis at 35 °C. Thus, under otherwise identical conditions, catalytic variation in the Rubiscos analysed is predicted to improve photosynthetic rates at physiological CO₂ concentrations. Naturally occurring Rubiscos with superior properties amongst the Triticeae tribe can be exploited to improve wheat photosynthesis and crop productivity.

Aegilops, barley, carboxylation, enzyme kinetics, photosynthesis, Rubisco, Triticeae, Triticum

Introduction

Wheat is the most widely grown crop and an important source of protein and calories, providing more than 20% of the calories consumed worldwide (Braun et al., 2010). It is projected that the world population will rise to over 9 billion by the year 2050 (United Nations, Department of Economic and Social Affairs, Population Division, 2013). This growth in population, along with a rise in per capita consumption (Kearney, 2010), will increase the global demand for food. Future increases in crop production will rely mainly on new strategies to increase yield and cropping intensity (Gregory and George, 2011; Alexandratos and Bruinsma, 2012; Fischer et al., 2014).

Yield traits that were positively affected by the green revolution appear to have relatively little remaining potential for further exploitation in modern wheat (Zhu et al., 2010), and further increases in yield potential will need to come from the improvement of photosynthetic efficiency. In this context, significant variation in biomass has been identified in exotic wheat genetic resources (Reynolds et al., 2015). Rubisco, (EC.4.1.1.39) is a key player in photosynthetic CO₂ assimilation, as it catalyses the first step of the Calvin–Benson cycle, fixing carbon dioxide through the carboxylation of ribulose-1,5-bisphosphate (RuBP). Rubisco also catalyses an additional and competing reaction with oxygen, which leads to the loss of fixed carbon and energy during photorespiration. This, together with the relatively low catalytic rate of Rubisco, limits photosynthetic productivity. Overcoming the limitations of Rubisco is therefore a major target in attempts to increase photosynthesis and yield (Parry et al., 2007; Parry et al., 2013).

There is natural variation in the catalytic properties of Rubisco isolated from various higher plants (Delgado et al., 1995; Galmés et al., 2005; Kapralov and Filatov, 2007; Andralojc et al., 2014; Galmés et al., 2014a; Galmés et al., 2014b). Relatively few studies report all catalytic parameters—the maximum velocities (V) and the Michaelis–Menten constants (K_M) for the carboxylase (c) and oxygenase (o) activities (V_c, V_o, K_c, and K_o, respectively) and the specificity factor (S_c/o=(V_c/K_c)/(V_o/K_o))—or measurements at anything other than a single temperature. Greater natural diversity is likely to be revealed when the catalytic parameters of Rubisco from a broader range of species become available. Current evidence suggests that there is a trade-off between the maximum carboxylation rate of Rubisco (V_c) and the relative specificity for CO₂ (S_c/o) (Bainbridge et al. 1995; Zhu et al., 2004; Savir et al., 2010), which may limit the extent to which these parameters can be independently altered. Clearly, a superior Rubisco for improving crop performance will have catalytic properties that maximize carboxylation, minimize the oxygenase activity, and enable enhanced rates of photosynthesis in relevant environments (Galmés et al., 2014b; Sharwood and Whitney, 2014).

Evolution of Rubisco variants with differing catalytic properties has been driven by their respective diverse cellular environments, which in turn are affected by their respective external environments (see Carmo-Silva et al., 2014 and references therein). These conditions provide selective pressures which favour changes in Rubisco structure that optimize performance (Tcherkez et al., 2006). While the chloroplast-encoded Rubisco large subunits incorporate the catalytic sites and therefore contribute directly to the observed catalytic properties, recent evidence suggests that changes in expression of genes within the nuclear-encoded small subunit multigene family can also cause catalytic variation (e.g. Morita et al., 2014).

The goal of this study was to identify Rubisco variants in the Triticeae tribe with catalytic properties that are likely to improve photosynthetic efficiency in wheat. We focused on wheat relatives so that useful traits could be introduced into a wheat genetic background by means of wide crossing (thus avoiding genetic manipulation), with increased likelihood that the available (wheat) chloroplast chaperones and Rubisco activase isoforms would subsequently promote the assembly and maintenance of catalytic activity in any resulting forms of the Rubisco holoenzyme. Triticeae genotypes from diverse climates and geographical locations were studied to increase the likelihood of identifying forms of Rubisco with different (and hopefully superior) kinetic properties from those found in Triticum aestivum (bread wheat). The resulting catalytic parameters were assessed in silico using a biochemical model of leaf photosynthesis. This approach suggested that Rubisco from two of the genotypes studied has the potential to improve the photosynthetic capacity and yield potential of wheat.

Materials and methods

Plant material and growth conditions

For all kinetic measurements, values obtained in test samples were compared with those of T. aestivum cv Cadenza, which was used as control. Cadenza is widely grown and has routinely been used in transformation experiments, making it a well-known and characterized variety. A total of 25 genotypes were analysed (Table 1). Species related to bread wheat were chosen with a range of characteristics, such as adaption to warmer conditions (T. aestivum SATYN and T. dicoccon CIMMYT), or which had been used to introduce desirable traits into bread wheat (Schneider et al., 2008). Seeds were obtained from CIMMYT (Mexico); the Royal Botanic Gardens, Kew (UK); and colleagues at Rothamsted Research. All plants were grown from seed in trays containing Rothamsted Research compost mix in a glasshouse at 20 °C with a 16h photoperiod. Additional lighting was provided whenever the photosynthetically active radiation (PAR) fell below 500 μmol m⁻² s⁻¹. All plants were well watered. Young, healthy leaves were harvested 2–3 weeks after sowing and rapidly frozen in liquid nitrogen.

Table 1.

Triticeae genotypes used to survey Rubisco catalytic properties for improving photosynthesis of UK bread wheat (T. aestivum cv Cadenza)

Haploid genome according to Van Slageren (1994). First letter denotes chloroplast genome.

Identity	Species name	Other species name(s) and additional information	Haploid genome
T. aestivum (C)	Triticum aestivum	Spring wheat var. Cadenza	BAuD
T. aestivum SATYN1	Triticum aestivum SATYN_II_9410	PUB94.15.1.12/FRTL (CIMMYT line)	BAuD
T. aestivum SATYN2	Triticum aestivum SATYN_II_9440	WHEAR//2*PRL/2*PASTOR (CIMMYT line)	BAuD
T. aestivum SATYN3	T. aestivum SATYN_II_9428	MTRWA92.161/PRINIA/5/SERI*3//RL6010/4*YR/3/PASTOR/4/ BAV92 (CIMMYT line)	BAuD
T. dicoccon1	Triticum dicoccon CI12214	Emmer wheat; INTRID:CWI47369 ENT:2129 (CIMMYT line)	BAu
T. dicoccon2	Triticum dicoccon CI12214	Emmer wheat; INTRID:CWI47368 ENT:2128 (CIMMYT line)	BAu
T. dicoccon3	Triticum dicoccon PI355483	Emmer wheat; INTRID:CWI45495 ENT:255 (CIMMYT line)	BAu
T. dicoccon4	Triticum dicoccon CI12214	Emmer wheat; INTRID:CWI47366 ENT:2126 (CIMMYT line)	BAu
T. timonovum	Triticum timonovum	Synthetic octoploid of T. timopheevii	GAm
T. timopheevii	Triticum timopheevii	Sanduri wheat	GAm
Triticale (Talentro)	Secale cereale × Triticum aestivum	× Triticosecale cv Talentro; frost and drought tolerant	BAuR
Triticale (Rotego)	Secale cereale × Triticum aestivum	× Triticosecale cv Rotego; frost and drought tolerant	BAuR
H. vulgare	Hordeum vulgare	Barley var. Lenins; relatively drought tolerant, not cold tolerant	H
Ae. tauschii	Aegilops tauschii	Aegilops squarrosa; drought tolerant	D
Ae. juvenalis	Aegilops juvenalis		DMU
Ae. vavilovii	Aegilops vavilovii	Drought tolerant	DMS
Ae. biuncialis	Aegilops biuncialis	Drought tolerant	UM
Ae. triuncialis	Aegilops triuncialis	Barbed goatgrass; Millenium Seed Bank 47689; winter annual	UC
Ae. comosa	Aegilops comosa		M
Ae. uniaristata	Aegilops uniaristata		N
S. cereale	Secale cereale	Rye var. Agronom; frost and drought tolerant	R
T. monococcum	Triticum monococcum	Millenium Seed Bank 11008, einkorn	Am
Ae. cylindrica	Aegilops cylindrica	Jointed goatgrass; cold tolerant	DC
Triticale (Cando)	Secale cereale × Triticum aestivum	× Triticosecale cv Cando; frost and drought tolerant	BAuR
Ae. speltoides	Aegilops speltoides	Not frost tender	S
B. distachyon	Brachypodium distachyon	Purple false brome; Accession BD21; diploid inbred

Specificity factor

Rubisco from snap-frozen young leaves (at least 500cm²) was extracted in homogenization buffer (40mM triethanolamine pH 8.0, 10mM MgCl₂, 0.5mM EDTA, 1mM KH₂PO₄, 1mM benzamidine, 5mM ε-aminocaproic acid, 50mM 2-mercaptoethanol, 5mM DTT, 10mM NaHCO₃, 1mM phenylmethanesulfonyl fluoride (PMSF), 1% w/v insoluble polyvinylpolypyrrolidone (PVPP)) at 0.3ml cm⁻². Leaves were homogenized in a pre-cooled blender for 45s and then filtered through four layers of muslin. Homogenate was clarified by centrifuging at 13 870×g for 12min at 4 °C, after which PEG₄₀₀₀ (60% w/v) was added to the supernatant to obtain a final concentration of 20.5% (w/v). MgCl₂ was added to the solution to increase the concentration of MgCl₂ to 20mM. The ensuing protein precipitation was complete after 30min at 0 °C, and the precipitate collected by centrifugation at 13 870×g for 20min at 4 °C. The protein precipitate was redissolved in column buffer containing 25mM TEA–HCl (pH 7.8), 5mM MgCl₂, 0.5mM EDTA, 1mM ɛ-aminocaproic acid, 1mM benzamidine, 12.5% glycerol, 2mM DTT and 5mM NaHCO₃. This was clarified by centrifugation at 175 000×g for 20min at 4 °C followed by filtration through a 0.45 µm regenerated cellulose syringe filter before further purification by anion-exchange chromatography on a 5ml HiTrap Q column (GE Healthcare, UK). Rubisco was eluted with a 0–1.0M linear NaCl gradient in the same buffer. Fractions with significant absorbance at 280nm were tested for Rubisco activity by measuring the RuBP-dependent incorporation of ¹⁴CO₂ into acid-stable products, as detailed below. Fractions showing Rubisco activity were pooled and further purified by size-exclusion chromatography on a Sephacryl S-200 column (GE Healthcare, UK) using a buffer consisting of 50mM Bicine–NaOH, pH 8.0, 10mM MgCl₂, 0.2mM EDTA, 10mM NaHCO₃ and 2mM DTT. Peak fractions based on Rubisco activity were pooled and concentrated using Pierce Protein Concentrators (150K MWCO, Thermo Scientific, UK). Samples were snap-frozen in liquid nitrogen and stored at –80 °C. Before use, samples were desalted by gel filtration through Sephadex G50 (medium grade; Sigma-Aldrich, UK) pre-equilibrated with assay buffer (0.1M Bicine–NaOH, pH 8.2, 20mM MgCl₂).

S_c/o was determined by measuring the decline in oxygen that accompanied the total consumption of RuBP in an oxygen electrode (Hansatech Instruments, UK), as described by Parry et al. (1989): Rubisco was activated in extracts by adding orthophosphate (4mM, pH 8.2) and NaHCO₃ (11mM), and incubating at 37 °C for 40min. A reaction mixture containing assay buffer and carbonic anhydrase (0.001% w/v, ≥2500 W-A units mg protein^–1; Sigma-Aldrich, UK) was equilibrated in an oxygen electrode vessel at controlled pH and temperature. All subsequent additions were made through a small aperture using glass syringes. Activated Rubisco and NaHCO₃ (2mM) were added to the vessel and the oxygen signal allowed to stabilize. RuBP (0.37mM) was added to the reaction, which was allowed to run to completion over a few minutes, as indicated by a stabilized oxygen signal. The amount of RuBP carboxylated was calculated by subtracting the oxygenated amount (represented by the amount of oxygen consumed during the reaction) from the amount added. The specificity factor was calculated as follows:

$$\begin{array}{l}{S}_{\text{c}/\text{o}}=\left(\text{RuBP carboxylated}/\text{RuBP oxygenated}\right)\\ \times \left({\text{O}}_{\text{2}}\text{concentration}/{\text{CO}}_{\text{2}}\text{concentration}\right)\end{array}$$

RuBP was prepared as previously described (Wong et al., 1980).

Rubisco catalytic properties

Rubisco was extracted from 20–30cm² of leaf material that was light-adapted immediately before being snap-frozen, then stored at –80 °C. Leaves were ground in an ice-cold mortar with 100mg quartz sand and 3.5ml of ice-cold extraction buffer, consisting of 100mM Bicine–NaOH, pH 7.9, 5mM MgCl₂, 1mM EDTA, 2mM benzamidine, 5mM ε-aminocaproic acid, 10mM NaHCO₃, 50mM 2-mercaptoethanol, 5% (w/v) PEG₄₀₀₀, 10mM DTT, 1% (v/v) plant protease inhibitor cocktail (Sigma-Aldrich, UK), 1mM PMSF and 2% (w/v) insoluble PVPP. After centrifugation for 5min at 14 000×g and 4 °C, samples were desalted by gel filtration through PD-10 columns (Sephadex G-25 Medium, GE Healthcare, UK) that had been pre-equilibrated with 100mM Bicine–NaOH, pH 8.0, 10mM MgCl₂, 1mM EDTA, 1mM benzamidine, 1mM ε-aminocaproic acid, 1mM KH₂P_i, 10mM NaHCO₃, 10mM DTT and 2% (w/v) PEG₄₀₀₀, 2% (v/v) protease inhibitor cocktail (Sigma-Aldrich, UK) and 20mM MgCl₂ was added before samples were snap-frozen and stored in liquid nitrogen, awaiting assay.

Catalytic parameters were measured essentially as previously described (Carmo-Silva et al., 2010). Carboxylation activity was measured at 8, 16, 24, 36, 68, and 100 µM CO₂ (aq) in equilibrium with a gas phase of N₂ supplemented with 0, 21, 60, or 100% (v/v) O₂. K_M and V_max for carboxylation (K_c and V_c, respectively) were calculated at each O₂ concentration using a Michaelis–Menten kinetic model. K_M and V_max for oxygenation (K_o and V_o, respectively) were calculated as follows:

$$K_{\text{o}}=\left[{\text{O}}_{\text{2}}\right]/\left[\left(K_{\text{M},\text{app}}/K_{\text{c}}\right)-\text{1}\right]$$

and

$$V_{\text{o}}=\left(V_{\text{c}}\times K_{\text{o}}\right)/\left(K_{\text{c}}\times S_{\text{c}/\text{o}}\right)$$

where K_c is the Michaelis–Menten constant for CO₂ in the absence of O₂, and K_M,app is the apparent Michaelis–Menten constant for CO₂ as measured in the reactions equilibrated with 21, 60, or 100% O₂. Specific mixtures of N₂ and O₂ were prepared using a gas divider (Signal Group, UK) and concentrations of O₂ in solution were calculated at 100% relative humidity and standard atmospheric pressure (101.3 kPa). At 25 °C, the solubility of O₂ was taken as 257.5 μM and the saturation vapour pressure of water as 11.6 kPa. At 35 °C, the solubility of O₂ was taken as 216.6 μM and the saturation vapour pressure of water as 12.0 kPa (http://www.eidusa.com/Theory_DO.htm). The concentration of CO₂ in solution (in equilibrium with HCO₃⁻) was calculated assuming a pK_a of 6.11 at 25 °C and a pK_a of 6.06 at 35 °C for carbonic acid, taking into consideration the pH of each buffer solution (measured on the day of assay). Carbonic anhydrase (≥77 WA units per 1ml reaction; Sigma-Aldrich, UK) was present in the reaction solution to maintain equilibrium between NaHCO₃ and CO₂. Control reactions were performed by measuring CO₂ fixation (acid-stable ¹⁴C) in reaction solutions lacking RuBP or NaHCO₃, as well as following total inhibition of Rubisco by prior treatment with an excess of the tight-binding inhibitor 2-carboxyarabinitol-1,5-bisphosphate (CABP). These controls confirmed that the activity measured was entirely due to Rubisco.

Radioactive content of ¹⁴C-labelled compounds was measured in 0.4–0.45ml aqueous solutions to which were added 3.6ml Ultima Gold Scintillation cocktail (Perkin-Elmer, UK), in a Tri-Carb 2100TR Liquid Scintillation Analyser (Perkin-Elmer, USA).

Turnover number (k_cat: mol product mol active site⁻¹ s⁻¹) was calculated from the corresponding V_max values (V_c and V_o: µmol acid-stable ¹⁴C mg Rubisco⁻¹ min⁻¹).

Rubisco quantification

Rubisco was quantified by the [¹⁴C]CABP binding assay described by Parry et al. (1997). For this, aliquots of the leaf extracts used in the assays described above, which had been snap-frozen immediately after extraction, were used. Each assay was performed in duplicate. Radioactive content of ¹⁴C-labelled compounds was measured as described above in ‘Rubisco catalytic properties’. Radiolabelled [2′-¹⁴C]CABP was prepared as previously described (Pierce et al., 1980).

Sequencing of Rubisco large subunit genes (rbcL)

Genomic DNA was extracted from young leaf tissue using the Qiagen DNEasy Plant Kit (Qiagen, UK). Partial rbcL fragments (equivalent to codons 1–463, c. 98% of the rbcL coding region) were amplified (Phusion HF polymerase, Invitrogen, USA) using the primers 5′rbcL_F2 (5′-TAATTCATGAGTTGTAGGGAGGG-3′) and cp063R (5′-TTTCCATACTTCACAAGCAGCAGCTAG-3′, from Dong et al., 2013), and cloned using the pGEM T-Vector Easy System (Promega, UK) with blue-white selection. For each genotype, multiple colonies with the fragment incorporated were identified and sequenced using the Eurofins Genomics service (Eurofins Genomics EU, Germany). Sequencing was performed using the primers M13 rev (5′-CAGGAAACAGCTATGACC-3′), M13 uni (5′-TGTAAAACGACGGCCAGT-3′), DRS15 (5′-CAAAAGTAGTAGAAACCATTTTAGTTCAGGTGG-3′ and DRS19 (5′-GKGYTCCTATTGTAATGCATGACTACTTAAC-3′). Sequence data were analysed using Geneious 7 (Biomatters; Kearse et al., 2012). Sequences obtained have been submitted to EMBL (http://www.ebi.ac.uk/ena/) and are publicly available (see Supplementary Table S1 at JXB online for accession numbers). Corresponding residue differences in the predicted large subunit (LSu) sequences appear in the format [Cadenza residue][residue position][test species residue] throughout the text.

Photosynthesis modelling

The effect of replacing native Rubisco in a wheat leaf with Rubisco from another species was modelled at 25 and 35 °C by entering the measured Rubisco catalytic constants into the biochemical models of carboxylation-limited and RuBP-limited C₃ photosynthesis (equations 2.20 and 2.23, respectively, in von Caemmerer (2000)). To accomplish this, values of K_c, K_o, and S_c/o were converted from units of concentration (mol l⁻¹) to those of partial pressure (bar), assuming solubilities of 3.34×10^–2 and 1.26×10^–3 mol (l bar)⁻¹ for CO₂ and O₂, respectively, for assays performed at 25 °C. At 35 °C, the respective solubilities were taken as 2.51×10^–2 and 1.083×10^–3 mol (l bar)⁻¹. We assigned a value of 38 μmol m⁻² for the estimated number of Rubisco active sites and kept this value constant for all samples. R_d was calculated as 0.015V_cmax. We assumed J_max as 1.5V_cmax at 25 ºC and 35 ºC giving a good fit above C_a. Equations used to generate the A–C_i curves were:

$$A_{\text{c}}=<\left[\left(C_{\text{i}}–\Gamma *\right)V_{\text{cmax}}\right]/\left\{C_{\text{i}}+\left[K_{\text{c}}\left(\text{1}+{\text{O}}_{\text{2}}/K_{\text{o}}\right)\right]\right\}>–R_{\text{d}}$$

and

$$A_{\text{j}}=\left\{\left[\left(C_{\text{i}}–\Gamma *\right)J_{\text{max}}\right]/\left(\text{4}{C}_{\text{i}}+\text{8}\Gamma *\right)\right\}–R_{\text{d}}$$

(von Caemmerer, 2000).

Statistical methods

Best-fit values of Michaelis–Menten constants (K_c and K_o) and maximum velocities (V_c and V_o) were derived from the kinetic data using Sigmaplot (v12.5). There was one determination per test genotype, with Cadenza values calculated from n=7 for catalytic properties and n=9 for S_c/o. Values of S_c/o at 25 °C were normalized to the corresponding value for the Rubisco of T. aestivum cv Cadenza (S_c/o=100), which was determined in parallel to each test sample measured (Parry et al., 1989). For S_c/o, the mean±SEM for every Rubisco preparation was calculated from a minimum of five technical replicates. Correlation coefficients were calculated using the Pearson product moment correlation test. The interaction between genotype and temperature was analysed using a non-parametric statistical approach. Ranking was done in descending order with the highest rank assigned number 1. Ranks of the measured variables for each genotype at 25 °C and 35 °C were correlated using Spearman’s rank correlation coefficient and these were tested for statistical significance using Genstat (17th edn, VSN International Ltd, Hemel Hempstead, UK).

Results

Rubisco catalytic properties at 25 and 35 °C were determined for 25 genotypes of Triticeae. For all genotypes, the maximum carboxylation velocity (V_c) was significantly higher at 35 °C than at 25 °C, ranging from 1.34 times higher in Ae. juvenalis to 2.65 times higher in T. dicoccon3 (Fig. 1 and Supplementary Tables S2 and S3). At 25 °C, H. vulgare ranked the highest for V_c, followed by Ae. cylindrica, T. aestivum SATYN3, and Triticale (Cando) above T. aestivum cv Cadenza (reference genotype). At 35 °C, T. dicoccon4 (a line developed by CIMMYT for warm climates) ranked the highest for V_c, followed by the other CIMMYT lines (T. aestivum SATYN and T. dicoccon genotypes) and H. vulgare, before Cadenza, which ranked seventh. At both temperatures B. distachyon ranked the lowest for V_c.

Fig. 1.

Rubisco carboxylation velocity (V_c) at 25 °C (black bars) and 35 °C (hatched bars) in 25 Triticeae genotypes. Data organized in decreasing rank at 25 °C, except for T. aestivum cv Cadenza, which is shown on the far left-hand side for comparison.

There is statistical evidence of a correlation between the performance of the genotypes with respect to V_c across temperature, with a Spearman’s rank correlation coefficient ρ=0.707 (P<0.001) (see Supplementary Tables S2 and S3). This value was slightly lower when comparing genotypes grouped according to rbcL sequence, although still significant (ρ=0.607, P=0.035) (Table 2).

Table 2.

Kinetic parameters of Rubisco at 25 and 35 °C according to the respective rbcL sequence (residues 1–463)

Where n>1, values are means±SEM for the Triticiae Rubiscos containing the same rbcL sequence. Other kinetic parameters are calculated using the Michaelis–Menten kinetic model as explained in text. For S_c/o, n≥5 technical replicates per genotype.

Assay temp	Representative genotype	n	Kc (µM)	Vc (µmol min–1 mg–1)	Ko (µM)	Vo (µmol min–1 mg–1)	Sc/o	kcat/Kc (21% O2)	Rank (Kc)	Rank (Vc)	Rank (Sc/o)
25 °C	T. aestivum cv Cadenza	12	16.3±0.4	3.01±0.03	431.6±12.3	0.85±0.02	95.94±0.96	0.14±0.00	6	4	7
	H. vulgare	1	15.2±2.4	3.47±0.16	465.3±27.8	1.09±0.14	101.96±5.21	0.17	3	1	4
	Aegilops spp.	9	15.6±1.0	2.63±0.08	431.4±17.6	0.74±0.02	100.84±2.20	0.13±0.01	4	6	5
	Ae. cylindrica	1	13.7±2.9	3.2±0.20	451.0±11.4	0.97±0.02	108.9±3.80	0.17	2	2	2
	Triticale (Cando)	1	16.1±2.4	3.15±0.14	384.0±4.7	0.75±0.01	99.73±5.08	0.14	5	3	6
	Ae. speltoides	1	16.5±3.2	2.82±0.17	446.9±11.0	0.84±0.02	102.3±5.6	0.13	7	5	3
	B. distachyon	1	11.9±2.5	1.78±0.10	395.7±18.8	0.54±0.03	111±4.00	0.1	1	7	1
35 °C	T. aestivum cv Cadenza	12	28.5±1.6	6.55±0.20	363.2±7.1	1.07±0.03	79.69±2.38	0.18±0.01	7	2	6
	H. vulgare	1	24.4±3.2	6.69±0.34	315.2±15.6	0.97±0.05	89.4±4.16	0.2	6	1	2
	Aegilops spp.	9	23.0±0.8	5.1±0.21	382.1±10.6	1.01±0.05	87.09±1.23	0.17±0.01	4	5	5
	Ae. cylindrica	1	20.7±3.7	4.91±0.32	310.1±2.8	0.83±0.01	89.1±1.60	0.17	2	6	3
	Triticale (Cando)	1	21.5±3.7	5.9±0.37	360.1±26.0	1.30±0.09	76.14±2.76	0.2	3	3	7
	Ae. speltoides	1	23.9±4.0	5.66±0.35	382.1±21.7	1.02±0.06	88.5±2.2	0.18	5	4	4
	B. distachyon	1	18.1±2.5	3.62±0.17	431.7±27.3	0.92±0.06	94±1.80	0.16	1	7	1

Rubisco from 14 genotypes had a higher affinity for CO₂ (lower K_c) than Cadenza at 25 °C, with B. distachyon ranking the highest (see Supplementary Table S2). Similarly, at 35 °C 13 genotypes showed a higher affinity for CO₂ (lower K_c) compared with Cadenza (Supplementary Table S3). Of these, only Rubisco from H. vulgare showed both a higher V_c and a higher affinity for CO₂ compared with Cadenza based on rank.

Regardless of the measurement temperature, Rubisco from most of the genotypes had a lower maximum oxygenation velocity (V_o) than Cadenza. Rubisco from genotypes that had a low affinity for O₂ (i.e. a high K_o) at 25 °C also showed a relatively low affinity for O₂ at 35 °C (Table 2 and Supplementary Tables S2 and S3).

Rubisco specificity factor (S_c/o) ranged from 90.4 (for Ae. juvenalis) to 111.0 (for B. distachyon) at 25 °C and was lower at 35 °C for all species (Fig. 2 and Supplementary Tables S2 and S3), ranging from 68.8 for T. aestivum SATYN1 to 94.0 for B. distachyon. In contrast to its ranking with respect to V_c, B. distachyon, ranked the highest for S_c/o at both temperatures. The CIMMYT lines T. aestivum SATYN3, T. dicoccon1, and T. dicoccon2 ranked much higher at 35 °C than at 25 °C with respect to S_c/o, although only T. dicoccon1 and T. dicoccon2 ranked higher than Cadenza. A correlation was identified in the performance of the genotypes across temperature with respect to S_c/o (ρ=0.857, P=0.003; Fig. 2 and Supplementary Tables S2 and S3). The correlation coefficient for S_c/o across temperature was lower but still significant when genotypes were grouped according to rbcL (ρ=0.503, P=0.002; Table 2). To compare S_c/o across temperatures, the respective ranks of individual genotypes were added up and compared (see Supplementary Tables S2 and S3). This revealed that B. distachyon (sum of ranks=2), Ae. tauschii (sum of ranks=8), T. monococcum (sum of ranks=8), Ae. cylindrica (sum of ranks=9) and Ae. triuncialis (sum of ranks=10) maintained their ranking much better than Cadenza (sum of ranks=28) across different temperatures.

Fig. 2.

Specificity factor of Rubisco (S_c/o) at 25 °C (black bars) and 35 °C (hatched bars) in 25 Triticeae genotypes. Data organized in decreasing rank at 25 °C, except for T. aestivum cv Cadenza, which is shown on the far left-hand side for comparison.

A positive correlation was observed between Rubisco V_c and K_c for all 25 genotypes, with this correlation being stronger at 35 °C (r=0.798, P<0.001) than at 25 °C (r=0.372, P=0.062) (Fig. 3). Rubisco from two genotypes (Ae. cylindrica and H. vulgare) appeared to have superior catalytic properties at 25 °C, possessing higher V_c and lower K_c values than Cadenza. From these, only Rubisco from H. vulgare retained superior properties at 35 °C compared with Cadenza, which performed remarkably well at this higher temperature. A strong positive correlation was found between V_c and V_o at 25 °C (r=0.726, P<0.001), while a moderate negative correlation was found between V_c and S_c/o at both temperatures (r=–0.428, P=0.029 at 25 °C and r=–0.528, P=0.006 at 35 °C).

Fig. 3.

Relationship between V_c and K_c for Rubisco from 25 Triticeae genotypes at 25 °C (A) and 35 °C (B) in the absence of O₂. Regression lines indicate the best fit through the data. Correlation coefficients (r) and P-values shown. The data point represented by Triticum aestivum cv Cadenza is highlighted by a circle. The area in the graph where Rubiscos with superior characteristics would be found is outlined. Arrows indicate genotypes with potentially superior Rubisco properties compared with Cadenza wheat. (i) H. vulgare; (ii) Ae. cylindrica.

Analysis of the rbcL coding sequences (codons 1–463) of the 25 genotypes revealed differences relative to the Cadenza reference sequence in 12 corresponding large subunit (LSu) residues, at positions spanning a number of domains within the Rubisco large subunit structure (Table 3). For 11 genotypes, all of which were either Triticum or Triticale, LSu sequences were identical to Cadenza at the amino acid level (Table 3 and Supplementary Table S1). Of the remaining 14 genotypes, each possessed at least one different amino acid from the Cadenza LSu. The LSu residue differences K14Q and S95N were the most common, and were found to occur together in all but one Aegilops species. These two residue differences were also found in S. cereale and T. monococcum. In H. vulgare, K14Q was the only difference relative to Cadenza rbcL. In Ae. cylindrica, in addition to K14Q and S95N, the LSu sequence also contained the difference V17A compared with Cadenza.

Table 3.

Amino acid differences in the Rubisco large subunit predicted protein sequences for 25 Triticeae genotypes relative to T. aestivum cv Cadenza

Residues under positive selection (Kapralov and Filatov, 2007, Galmés et al. 2014b) are indicated with an asterisk. Functional interactions described in the literature for these residues as indicated (AS, active site; ID, intradimer interactions; DD, dimer:dimer interactions; RA, interactions with Rubisco activase; SSU, interaction with small subunits). Symbols and colours match those used in Fig. 4. na, not applicable.

Residue change	Symbol	Interaction	Location of residue	Species
na				T. aestivum cv. Cadenza
na				T. aestivum SATYN1
				T. aestivum SATYN2
				T. aestivum SATYN3
				T. dicoccon1
				T. dicoccon2
				T. dicoccon3
				T. dicoccon4
				T. timonovum
				T. timopheevii
				Triticale (Talentro)
				Triticale (Rotego)
K14Q*			N-terminal	H. vulgare cv. Lenins
K14Q*			N-terminal	Ae. tauschii
S95N*		ID, RA		Ae. juvenalis
				Ae. vavilovii
				Ae. biuncialis
				Ae. triuncialis
				Ae. comosa
				Ae. uniaristata
				S. cereale cv. Agronom
				T. monococcum
K14Q*			N-terminal	Ae. cylindrica
V17A			N-terminal
S95N*
G47W		ID	Strand B	Triticale (Cando)
K81R				Ae. speltoides
I225T*		SSU	Helix 2
G10S			N-terminal	B. distachyon
K21R			N-terminal
A91P*		RA
I251M*		DD, ID, SSU	Helix 3
S328A*		AS	Loop 6
M341I		AS, ID	Helix 6

The catalytic efficiency of Rubisco in air (21% O₂) can be measured as the ratio between the carboxylase turnover number and the Michaelis–Menten constant for CO₂ (i.e. k_cat/K_c at 21% O₂, Table 2). Rubisco from Ae. cylindrica and H. vulgare appeared to have a superior efficiency to Cadenza (and other genotypes with the reference LSu sequence) at 25 °C. Rubisco from H. vulgare also showed superior efficiency at 35 °C, along with Triticale (Cando).

From the relationship between catalytic efficiency (k_cat/K_c at 21% O₂) and S_c/o it is also possible to identify Rubisco enzymes with superior catalytic performance (Fig. 4). In general, catalytic efficiency was higher, but S_c/o was lower, at 35 °C compared with 25 °C. Forms of Rubisco with the same LSu sequence (residues 1–463) showed considerable variation in their combination of catalytic properties. Rubisco from H. vulgare, which only differs from Cadenza by virtue of the alternative residue Q14 (K14 in Cadenza), stood out as having a promising combination of k_cat/K_c and S_c/o at both temperatures. Rubisco from Ae. cylindrica, differing by K14Q, V17A and S95N relative to Cadenza, showed promise only at 25 °C. Rubisco from Ae. vavilovii showed a similar catalytic response, despite the LSu sequence being identical to that from a number of other species that did not show such catalytic advantage compared with Cadenza. In B. distachyon, six residue changes compared with Cadenza (Table 3) might be associated with a higher S_c/o, but at the expense of catalytic efficiency (Fig. 4).

Fig. 4.

The relationship between the catalytic efficiency of Rubisco at 21% O₂ (k_cat/K_c, µM s⁻¹) and the specificity factor (S_c/o) of Rubisco at 25 °C (circles) and 35 °C (triangles). Each colour denotes an rbcL sequence (as per Table 3) and Cadenza wheat (C, used as reference) is represented by the diamond and square at 25 and 35 °C, respectively.

The Rubisco catalytic constants measured in vitro at 25 and 35 °C for Cadenza, Ae. cylindrica and H. vulgare were used to assess the theoretical impact on photosynthetic performance of wheat leaves over a range of intercellular CO₂ concentrations, including those corresponding to ambient air (c. 210 µbar). This modelling exercise predicted that at 25 °C photosynthetic rates would be improved by replacing wheat Rubisco with the enzyme from either Ae. cylindrica or H. vulgare (Fig. 5A, B). At 25 °C Rubisco from Ae. cylindrica showed a maximal increase in assimilation rate of 23% (6.3 µmol m⁻² s⁻¹) compared with Cadenza at 270 μbar C_i (Fig. 5A), while Rubisco from H. vulgare maximally increased assimilation rate by 22% (6.7 µmol m⁻² s⁻¹) at 300 μbar C_i (Fig. 5B). Rubisco from H. vulgare also showed promise for the improvement of photosynthesis in wheat at 35 °C, while the enzyme from Ae. cylindrica was inferior to Cadenza wheat at the higher temperature (Fig. 5C, D).

Fig. 5.

Modelling photosynthesis at 25 °C (A, B) and 35 °C (C, D), to demonstrate the benefit of replacing Rubisco of T. aestivum cv Cadenza (red) with Rubisco from Ae. cylindrica (A, C; blue) or H. vulgare (B, D; blue). Rubisco-limited (A_c, solid lines) and RuBP regeneration-limited (A_j, dashed lines) rates of net CO₂ assimilation (A) were derived using the model of Farquhar et al. (1980) and the Rubisco catalytic constants measured in vitro for each genotype. Blue shading indicates where Rubisco from the test genotypes showed higher assimilation rates than native Cadenza Rubisco.

Discussion

The catalytic properties and primary sequence of the Rubisco large subunits (LSu, encoded by rbcL) from 25 Triticeae genotypes revealed diversity relevant to improving wheat photosynthetic performance in current and projected warmer temperatures. In the major wheat producing countries, grain filling is accompanied by increasing daytime temperatures (Asseng et al., 2015). Within the limits of resources available to this study, measurements were taken at an ideal growth temperature (25 °C, Nagai and Makino, 2009), and at an elevated temperature at which a pronounced negative impact on yield would be expected (35 °C, Duncan et al., 2014). At the higher temperature V_c was higher, but S_c/o was lower, which is consistent with previous research (Brooks and Farquhar, 1985; Tcherkez et al., 2006; Savir et al., 2010; Galmés et al., 2014b). Differences were observed in the Rubisco response to temperature that suggests some acclimation to different geographical locations.

As reported previously (Savir et al., 2010), a positive correlation between V_c and K_c, the determinants of Rubisco carboxylase efficiency, was observed at both temperatures for the Triticeae genotypes studied here, indicating that genotypes with a high V_c tend to have lower affinity for CO₂. The catalytic efficiency of Rubisco in air was a useful tool to identify Rubiscos with superior performance. Furthermore, the combined results suggest that all of the Rubisco catalytic properties, including the specificity factor, must be taken into account during the search for forms of Rubisco with improved performance in air. This follows from the parameters required for biochemical modelling of photosynthetic performance, which include V_c, K_c, and S_c/o (=V_c.K_o/V_o.K_c), the latter being used to determine the compensation point (Γ*=0.5[O₂]/S_c/o) in the absence of dark respiration (von Caemmerer, 2000).

In wheat, variation in V_c has been observed across different genotypes and it has been suggested that many of the catalytic properties of Rubisco are determined by the large subunit (Evans and Austin, 1986; Terachi et al., 1987; Kasai et al., 1997), which contains the catalytic sites. The rbcL gene is chloroplast encoded (Spreitzer and Salvucci, 2002) and the chloroplast genome tends to be evolutionarily highly conserved. However, within the Poaceae, rbcL has evolved at a relatively rapid rate compared with other families of flowering plants (Bousquet et al., 1992; Gaut et al., 1992). Since the large subunits contribute directly to catalytic function, variation in this sequence in wheat relatives represents a potential source of improved catalytic activity.

The majority of the Triticeae genotypes characterized in this study are highly inter-related and this was reflected in the similarity of the respective rbcL sequences. Some of the observed differences in Rubisco catalytic activity correlated with differences in rbcL sequence. For example, differences in catalytic properties determined for H. vulgare, Ae. cylindrica, Triticale (Cando) and B. distachyon compared with Cadenza Rubisco might be associated with their specific rbcL sequences. Conversely, differences in catalysis for genotypes with the same rbcL sequence (e.g. the Cadenza group represented by black symbols or the Aegilops group represented by yellow symbols in Fig. 4) may be associated with either changes in the extreme C-terminus whose sequence was not determined or, more likely, diversity in the small subunit sequence (Guo et al., 1997; Spreitzer, 2003; Spreitzer et al., 2005; Ishikawa et al., 2011; Cai et al., 2014; Morita et al., 2014). Future studies to characterize the exact number and relative expression of small subunit genes in wheat and wheat relatives may reveal novel avenues for improving Rubisco catalysis and photosynthesis.

When comparing kinetic parameters between species grouped by LSu sequence (Table 2), H. vulgare (difference K14Q) ranked highest in V_c at both 25 and 35 °C, with only Ae. cylindrica (K14Q, V17A, and S95N) also ranking higher than species with the control sequence at 25 °C. Rubisco from Ae. cylindrica also had a lower K_c (higher affinity for CO₂) at both temperatures compared with Rubisco from species with the reference rbcL sequence.

When present as lysine, large subunit residue 14 is known to be a site of post-translational tri-methylation in many flowering plant species, but not T. aestivum (Houtz et al., 2008). Available data show that glutamine is the only alternative residue found at this position (K14Q, Houtz et al., 1989; Houtz et al., 1992; Trievel et al., 2003), although the importance of this position and its modification remains unresolved (Houtz et al., 2008). The results presented here suggest the possibility that either the amino acid difference itself (K14Q) or the absence of methylation at this position alters Rubisco kinetics in a manner favourable to photosynthesis.

The other residue difference common to most of the Aegilops, S. cereale and T. monococum, S95N, occurs in a poorly conserved region of the rbcL gene, which is in the proximity of residues known to be involved in interactions with Rubisco activase (Portis, 2003; Portis et al., 2008; Carmo-Silva et al., 2014). Recently, this residue was highlighted during a search for residues under positive selection, and it is in proximity to residues involved in L₂ intradimer interactions (Galmés et al., 2014b). Interestingly, species combining the K14Q and S95N residue differences (but having no other differences) showed no consistent catalytic difference compared with the reference LSu sequence group.

The valine at position 17 is known to be involved in intradimer interactions (Knight et al., 1990; Kellogg and Juliano, 1997). Rubisco from Ae. cylindrica containing V17A in combination with K14Q and S95N had improved carboxylation catalytic efficiency at 21% O₂ and 25 °C compared with the reference Cadenza (Table 2). This form of Rubisco was predicted by modelling to improve photosynthetic performance at all CO₂ levels relative to Cadenza at 25 °C (Fig. 5A). Hence, the influence exerted by the relatively conservative Val–Ala (V17A) difference, when combined with K14Q (and S95N), may explain the positive impact on Rubisco catalysis.

H. vulgare Rubisco had improved k_cat/K_c compared with Cadenza, while only containing the K14Q difference. Studies in Anacystis nidulans found that a K14Q or K14L mutation had no influence on enzyme activity (Kettleborough et al., 1991). While the present study did not cover the extreme C-terminus of the large subunit, available sequences showed a KV extension in that region of the H. vulgare sequence (Petersen and Seberg, 2003), which may be relevant to its superior catalysis in comparison to Cadenza. Confirmation of this hypothesis would be valuable, given that modelling of the photosynthetic response to intercellular CO₂ predicts a benefit at both 25 and 35 °C by replacing the native Rubisco with the barley enzyme.

As with Ae. cylindrica (Colmer et al., 2006), barley is considered to be a valuable genetic resource for improving stress tolerance in wheat (Dulai et al., 2011; Molnar-Lang et al., 2014). Barley–wheat hybrids have been investigated before (Kruse, 1973; Malik et al., 2011; Dulai et al., 2011; Rodríguez-Suárez et al., 2011; Pershina et al., 2012; Zou et al., 2012), but without a focus on yield improvement. One notable exception is the development of Tritordeum, which is a hybrid between wild barley (H. chilense) and durum wheat (T. turgidum ssp. Durum, haploid genome BA; Martin et al. 1996), which has been commercialized (http://www.agrasys.es/). Tritordeum has particularly high protein content and has shown tolerance to drought conditions in field trials (Martin et al., 1999; Villegas et al., 2010). While Tritordeum does not include the D genome present in T. aestivum, data in this study suggest that this hybrid warrants further investigation with respect to Rubisco kinetics and yield potential.

This study has identified residues that warrant further study, e.g. by mutagenesis. Well targeted single amino acid changes can have a dramatic impact on catalytic performance (e.g. Whitney et al., 2011). However, at present there is no available expression system to test the effect of amino acid substitutions on Rubisco from monocots. An alternative, and possibly more promising approach, which utilizes available technology, is the introgression of traits through wide-crossing of Triticeae genotypes.

Conclusion

The Rubisco catalytic properties determined for 25 genotypes showed that variation exists even amongst closely related genotypes. Rubisco from Ae. cylindrica and H. vulgare showed promising catalytic properties that should be explored in the context of improving photosynthesis, and ultimately yield, in wheat. Ideally, this could be carried out by crossing a number of the species examined here with bread wheat and studying the resulting plants with respect to Rubisco catalytic activity, photosynthesis and yield. This study supports the case for investment in genetic resource screening for photosynthesis-related characteristics.

Supplementary data

Supplementary data are available at JXB online.

Table S1. Rubisco large subunit (rbcL) single nucleotide polymorphisms.

Table S2. Rubisco catalytic parameters at 25 °C.

Table S3. Rubisco catalytic parameters at 35 °C.

Acknowledgements

This work was funded by CIMMYT (W4031.11 Global Wheat Program) and by a subcontract to the University of Illinois as part of the Bill and Melinda Gates Foundation award RIPE, Realizing Increased Photosynthetic Efficiency (rbcL sequencing work). MAJP, PJA and EC-S acknowledge support from the Biotechnology and Biological Sciences Research Council through the 20:20 Wheat^® Institute Strategic Program (BBSRC BB/J/00426X/1). The authors are grateful to Spencer Whitney (Australian National University) and Pippa Madgwick (Rothamsted Research) for useful discussions on rbcL sequencing, to Simona Vătavu (Rothamsted Research) for technical assistance, and to Stephen Powers for support with statistical analysis. Seeds were kindly provided by the Millennium Seed Bank, Kew Gardens, and by colleagues at Rothamsted Research (Dr Mark Wilkinson, Dr Caroline Sparks, Dr Jennifer Postles, Dr Till Pellny and Mr Steve Harvey).

Author notes

*

Present address: Department of Natural Sciences, Middlesex University, The Borroughs, London NW4 4BT, UK

‡

Present address: Lancaster Environment Centre (LEC), Lancaster University, Lancaster LA1 4YQ, UK

Editor: Christine Raines, University of Essex