Due to the growing world population, crop yields must increase to meet the rising demand. Crop plants also require adaptation to optimize performance in the changing environments caused by climate change. Improving photosynthetic carbon fixation is a promising, albeit technically challenging, strategy whose potential has only just begun to be considered in breeding programmes. Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco), a fundamental enzyme of carbon fixation, is extremely inefficient and many strategies to improve photosynthesis focus on overcoming the limitations of this enzyme, either by improving Rubisco activity and regulation or by improving the supply of substrates. Although progress is being made, the need to tailor solutions for each crop and their respective environments has been highlighted. Even so, continuing research will be required to achieve these objectives and to grow crops more sustainably in the future.
Photosynthetic CO2 fixation in the Calvin–Benson–Bassham (CBB) cycle is the source of almost all the organic carbon in the biosphere. There is compelling evidence from a plethora of free air CO2 enrichment experiments that increasing CO2 fixation will increase crop biomass and yield . The experimental evidence that increasing photosynthesis will increase crop yields is reinforced by modelling studies [2–4]. Increasing photosynthetic CO2 fixation is widely recognized as a key objective in efforts to meet the demand for increased crop production and to secure global food security. Despite this, there have been very few attempts to consider, let alone incorporate, photosynthetic traits in breeding programmes [5–7]. Wider recognition of the potential benefits have recently resulted in several major international initiatives, focused on increasing photosynthetic CO2 fixation in specific crops (e.g. C4 rice, https://c4rice.com/; RIPE, http://ripe.illinois.edu/; International Wheat Yield Partnership, http://iwyp.org/). Improving carbon fixation has been the subject of a considerable number of detailed reviews (e.g. [8–11]). Here, we focus on the most recent developments in this area.
The rate of the CBB cycle is co-limited by the regeneration of the CO2 acceptor, ribulose-1,5-bisphosphate (RuBP) and by the reaction of CO2 with RuBP, catalysed by RuBP carboxylase/oxygenase (Rubisco, EC 22.214.171.124). The extent to which each of these processes limits the rate of CO2 assimilation is dependent on the relative abundance and activity of the underpinning molecular machinery, which is strongly influenced by environmental conditions.
Overcoming the inefficiencies of Rubisco continues to be a major focus of research (Figure 1). In addition to catalysing the productive carboxylation of RuBP in carbon assimilation, Rubisco also catalyses a competing and wasteful reaction with oxygen which necessitates the operation of the photorespiratory cycle, to return the diverted carbon to the CBB cycle. In addition, Rubisco from higher plants (and particularly those which lack a carbon concentrating mechanism (CCM)) conducts successive rounds of catalysis (or turnover) at relatively low rates compared with the other enzymes of the CBB cycle. This necessitates the synthesis of large amounts of the enzyme (accounting for 30–50% of the soluble protein) in crop leaves , in order to sustain the observed photosynthetic rates. Furthermore, the complex reaction mechanism of Rubisco results in not infrequent abortive side reactions, causing the production of tight binding inhibitors that prevent further catalysis. Despite these inefficiencies, Rubisco remains the only enzyme able to deliver the net assimilation of carbon on a global scale. Some alternative natural and synthetic carbon fixation pathways have been identified  and may be useful in future approaches to increase carbon fixation. As elaborated below, research into overcoming the limitations to photosynthetic CO2 assimilation has focused on both improving the supply of CO2 and RuBP to Rubisco, as well as improving the efficiency, speed and regulation of this enzyme.
Strategies to improve carbon fixation based on three key processes: increasing CO2 supply, Rubisco and its regulation, and RuBP supply
To become catalytically competent, the catalytic sites of Rubisco must assume a specific active site geometry. In this process of activation—which is reversible—the carbamylation of an essential lysine residue within the catalytic site, followed by stabilization of the resulting carbamate by a Mg2+ ion  gives rise to a catalytically active ternary complex (E.CO2.Mg2+). The carboxylase and oxygenase activities of Rubisco both occur at this activated catalytic site. Neither gaseous substrate binds directly to Rubisco prior to catalysis, but they compete for reaction with an enzyme-bound enediol isomer of RuBP. The chemistry of the carboxylase mechanism is reasonably well understood and entails a sequence of five partial reactions: enolization, carboxylation, hydration, carbon–carbon cleavage and protonation with a stereochemical inversion resulting in the two molecules of d-3-phosphoglycerate (3PGA, ). However, some uncertainty remains as to whether the chemical mechanism is stepwise or concerted . The oxygenase activity is initiated by the reaction between oxygen and the enediol of RuBP, and is followed by hydration and carbon–carbon cleavage to produce 2-phosphoglycolate (2PGA) and 3PGA  following a reaction pathway which is likely to be singlet O2-independent  and so unlikely to play a role in the removal of reactive oxygen species. Rubisco also catalyses periodic ‘misfire’ reactions, in which a peroxy-intermediate of the oxygenase pathway decomposes to yield hydrogen peroxide and pentodiulose-1,5-bisphosphate (PDBP, ). PDBP can further rearrange to 2-carboxytetritol 1,4-bisphosphate . Both these bisphosphate compounds can bind with high affinity to activated Rubisco, preventing CO2 fixation . Furthermore, mis-protonation of the enediol of Rubisco-bound RuBP also occurs, forming d-xylulose-1,5-bisphosphate (XuBP, ) which is another potent, competitive inhibitor of catalysis. It was recently shown that a ubiquitous CbbY protein possessed XuBP phosphatase activity, with the potential to eliminate such inhibition and the consequent efficiency penalty .
The high degree of amino acid sequence and structural homology between forms of Rubisco from diverse species, imply a conserved catalytic mechanism . Even so, significant interspecies variation in the magnitude of common catalytic parameters has been reported [24,25]. Despite its abundance in planta and the vast database of Rubisco sequence information spanning the plant kingdom, the extent to which the associated catalytic parameters vary per se has only just begun to be addressed. Until recently, our knowledge of the catalytic variability of Rubisco was comparatively small, because only a few species had been examined and even among these a complete kinetic characterization—including rate constants and substrate affinities for both carboxylase and oxygenase activities (kcatc, kcato, Kc and Ko respectively) as well as estimates of Rubisco abundance—was rare. The catalytic properties of Rubisco isolated from this very small number of species had suggested an inverse relationship  between the carboxylase capacity (kcatc) of Rubisco and the corresponding specificity factor (Sc/o), which reflects the ability of the enzyme to discriminate between CO2 and O2 . This in turn led to the hypothesis that there was an obligatory trade-off between optimization for speed on one hand and optimization for substrate affinity on the other , and thence to the suggestion that Rubisco may already be optimized to fulfil its role in any specific photosynthetic organism  on account of the prevailing evolutionary constraints . Since then, Rubisco has been extracted from a much larger and more diverse variety of sources and the kinetic properties fully determined, often over a range of temperatures [25,31–34]. All these studies have demonstrated significant variability in the kinetic properties of Rubisco from different species. Importantly, no correlation between kcatc and Sc/o (inverse or otherwise) was detected among the 11 types of diatom Rubisco of Form ID (or ‘red’-type lineage) which were studied by Young et al. . Even more surprising were the highly significant positive correlations between kcatc and Sc/o demonstrated between Form IB Rubisco from 75 diverse plant species at 20 and 25°C . Together, these more comprehensive studies demonstrate that any evolutionary constraints imposed by an inverse relationship between the speed and affinity of Rubisco, have important exceptions, giving renewed hope that both might be favourably engineered within the same holoenzyme.
The observed variations in kinetic properties probably arise from differences in the enolization equilibrium and a change in the energy barriers associated with CO2 and O2 addition. But how specific amino acid sequences determine this variation remains unclear . Sequence analysis of many of the species investigated, has identified some amino acid residues that alter catalytic activity and could be targeted in attempts to engineer improved Rubisco efficiency in crop systems [25,36].
In higher plants, the (Form I) Rubisco holoenzyme is composed of eight chloroplast-encoded large subunits (L) arranged as an L8 (4 × L2) core, which is capped at opposite, structurally equivalent, poles by two tetrads of nuclear-encoded small subunits (2 × S4) . While the large subunits contain the catalytic sites and are the primary determinants of the catalytic properties of Rubisco, there is evidence that Rubisco kinetic properties are also affected by the small subunits [38,39]. Within a species, the L subunits of Rubisco have identical amino acid sequences, but the small subunits are encoded by a multigene family for which individual members are differentially expressed. The S subunit gene sequences within this family are generally so highly conserved that the amino acid sequences of the S subunits they encode are identical (Spreitzer et al. (2003) ). However, a distinct S subunit gene sequence with lower homology, present on a distinct chromosome and encoding a distinct S subunit, was recently demonstrated in rice. The expression of this divergent sequence was tissue specific and the corresponding Rubisco holoenzyme had distinct catalytic properties .
Integral to the biochemical models of photosynthetic CO2 assimilation of Farquhar et al.  are the relative contributions of RuBP regeneration and the role played by Rubisco in the rate of CO2 assimilation. Specifically, the way in which Rubisco abundance, carboxylase capacity, activation state, specificity and affinity for oxygen, impact upon CO2 assimilation are all taken into account, making the model a powerful predictive tool. In this way, the theoretical outcome of substituting a native Rubisco with that from other plant species can be assessed across a range of CO2 concentrations. Modelling indicates that the direct replacement of native Rubisco from major crops with specific high-performing Rubiscos, would support significant increases in leaf photosynthesis at current atmospheric CO2 levels and high irradiance [2,10]. Typically, the predicted improvements were most pronounced at relatively low internal CO2 concentrations, where leaf photosynthesis is typically limited by Rubisco activity; this may be advantageous under conditions where water is limiting, since attempts to reduce water loss by stomatal closure would result in lowered internal CO2 concentrations [25,32,43]. However, the extent of the predicted improvement was found to depend on the specific crops and their growth environment. In addition, as the Rubisco kinetic constants also have different thermal responses, it has become apparent that any engineering solutions will need to be tailored to each crop and environment [25,32–34].
Although it may be possible to introduce novel Rubiscos by conventional breeding (e.g. wide crosses ), in most cases a biotechnological approach appears essential if the benefits of the most promising (and more distantly related) forms of Rubisco are to be realized. Thus far, attempts to introduce novel forms of Rubisco by nuclear transformation have had limited success. In contrast, the introduction of novel Rubisco via plastid transformation has been much more successful [24,36,43,44] although this technology is not yet available for any of the major food crops (i.e. maize, wheat and rice). Optimizing the catalytic properties via the Rubisco S subunit offers a potentially exciting alternative or complementary avenue to enhance crop photosynthesis . However, even when plastid transformation becomes more widely applicable, it cannot be assumed that any non-native or modified Rubisco would be assembled and be functional in planta. When novel Rubiscos have been expressed in tobacco, only those from higher plant species most closely related to tobacco are functional. Although it has been possible to express functional cyanobacterial Rubisco in transplastomic tobacco , the amount of functional enzyme was low . In general, Rubisco from more distantly related species (e.g. green algae and monocots with respect to a dicot recipient) is not assembled into functional enzyme [24,36] due to chaperone incompatibilities .
The activity of Rubisco in vivo is determined by carbamylation and/or by the tight binding of inhibitors . In daylight, the catalytic misfire product PDBP and other inhibitors can lock active sites into an unproductive form . In some species during periods of low light or darkness 2-carboxy-d-arabitinol 1-phosphate (CA1P), a naturally occurring transition state analogue of the carboxylase reaction is synthesized and binds tightly to Rubisco, preventing catalysis . While inhibitors prevent catalysis, they do also protect Rubisco from proteolytic breakdown and oxidative damage under stress [20,48]. The removal of such inhibitors from Rubisco requires the action of the ancillary protein Rubisco activase (Rca, EC 126.96.36.199) in an ATP-dependent manner.
Rca is a member of the AAA+ super family whose members perform chaperone-like functions . The rate of inhibitor removal by Rca determines the duration of photosynthetic induction, i.e. the time required to reach a constant steady-state rate of net CO2 assimilation upon transition from low to high irradiance [50–54]. This is especially important considering that plants in an agronomic context will frequently experience fluctuations in the light environment and rarely operate under steady-state conditions of photosynthesis. Taylor and Long  have recently demonstrated in wheat that the lag in photosynthetic induction associated with Rubisco activation during shade-to-sun transitions could diminish the carbon assimilation potential by 21%. While this lag in photosynthetic induction occurs in most plant species studied to date , in Arabidopsis plants engineered to express only a non-regulated form of Rca, photosynthetic induction in response to increases in irradiance was faster than in wild-type (wt), and biomass accumulation was enhanced relative to wt when grown under a periodically fluctuating irradiance regime . These observations make it realistic to explore variation in the response of Rubisco activation to irradiance increases when attempting to optimize the efficiency of carbon assimilation under agronomic environments characterized by fluctuations in irradiance.
The ubiquitous presence of promiscuous sugar phosphatases, such as CA1P and XuBP phosphatases, which dephosphorylate a range of Rubisco inhibitors rendering them non-inhibitory [22,56] suggests a role for Rubisco inhibition in nature. The potential for Rubisco inhibition to protect Rubisco from proteolytic breakdown under conditions of environmental stress [20,48] remains to be tested in planta. However, in agronomical settings where attempts are made to reduce the effects of stress, there is potential to realize increased biomass production by accelerating the adjustment of photosynthetic CO2 assimilation in shade-to-sun transitions, as recently demonstrated in field-grown tobacco engineered to adjust more rapidly during shade-to-sun transitions .
Increasing RuBP supply
The regeneration of RuBP co-limits the CBB cycle and the share of this control is predicted to increase as atmospheric CO2 concentrations continue to rise. Modelling studies  have suggested that the CBB cycle is not optimized, but that overexpression of the CBB cycle enzymes: sedoheptulose-1,7-bisphosphatase (SBPase: EC.188.8.131.52), fructose-1,6-bisphosphate aldolase (FBPA: EC 184.108.40.206) and Rubisco would increase carbon assimilation. However, in most cases, any such optimization would need to be tailored to the specific crop and growth environment and would itself require further modification in response to continuing climate change .
SBPase catalyses the dephosphorylation of sedoheptulose-1,7-biphosphate (SBP) to sedoheptulose-7-phospate in the first committed step of RuBP regeneration. SBPase alone exerts substantial control over RuBP regeneration. Glasshouse experiments with transgenic plants have confirmed that overexpression of SBPase can increase photosynthetic rate and biomass in several species: tobacco , rice  and wheat , although the magnitude of the response was variable. However, in field trials with the same transgenic rice lines, increases in biomass were only seen at elevated CO2 concentrations or under temperature stress .
FBPA promotes the condensation of dihydroxyacetone phosphate (DHAP) and glyceraldehyde-3-phosphate (G3P), forming fructose-1,6-bisphosphate (F1,6P). Glasshouse experiments with transgenic tobacco plants have confirmed that overexpression of FBPA can increase photosynthetic rate and biomass in tobacco, although the stimulation was greatest when the plants were grown with variable square wave lighting and at elevated CO2 concentrations, rather than at ambient CO2 .
In transgenic tobacco plants, the overexpression of SBPase and FBPA together increased the CO2 assimilation rate at intercellular CO2 concentrations above 300 μmol.mol–1 compared with both the non-transformed controls and to lines overexpressing the SBPase alone . At ambient CO2, plants overexpressing SBPase and FBPA also had a significantly higher light saturated rate of photosynthesis than the wt. Overexpression of FBPA together with SBPase had a positive effect on growth in high light when compared with either non-transformed plants or plants overexpressing SBPase alone, but this cumulative effect of co-expression did not occur when plants were grown in low light .
The 2PGA generated by the oxygenase activity of Rubisco is converted into 3PGA during photorespiration (two molecules of 2PGA are required to form one 3PGA), and subsequently into RuBP by the CBB cycle. A pivotal, multifunctional enzyme complex of photorespiration is mitochondrial glycine decarboxylase (GDC). Overexpression of the H protein of GDC in transgenic Arabidopsis plants, increased both net carbon fixation and growth . This result contradicted earlier models, which had predicted that at current atmospheric CO2 concentrations there was an overinvestment in the enzymes of this cycle . The results of Timm et al. , indicate that the apparent overinvestment may actually have been due to an underinvestment in this CN carrier (the H-protein). Simkin et al.  recently demonstrated that overexpression of GDC-H in tobacco in combination with SBPase and FBPA overexpression, resulted in cumulative positive effects on leaf area and biomass accumulation. The value of such multigene manipulations in plants grown under field conditions awaits demonstration , although this technology has the potential to deliver greater benefits than single gene approaches.
An alternative approach has been to diminish the penalties (i.e. energy consumed and loss of fixed carbon and nitrogen) of photorespiration by modifying the photorespiratory cycle. More efficient alternative metabolic pathways that incorporate enzyme-catalysed reactions only found in bacteria, algae and Archeae and which complement those already present in higher plants have been designed and tested. While some did not appear effective , others were accompanied by increases in biomass [68,69]. Modelling studies have predicted that the impact of the photorespiratory bypasses vary widely depending on the species and the growth environment [70,71]. The value of such manipulations for the growth of crop plants under field conditions must now be confirmed.
Increasing the concentration of CO2 at the catalytic site of Rubisco will increase the proportion of carboxylation relative to oxygenation, thereby increasing overall carbon assimilation and biomass production. This has been achieved in some species (e.g. C4 plants, CAM plants) through the evolution of CCMs. One approach to increasing photosynthesis in C3 crop plants such as rice or wheat, would be the introduction of one of these mechanisms.
As C4 photosynthesis has evolved independently many times before [72,73], it may be possible to introduce a functional C4 pathway into crops that lack one. In the C4 pathway, the initial fixation of CO2 catalysed by mesophyll phosphoenolpyruvate carboxylase, is followed by product translocation and decarboxylation to release CO2 in the vicinity of bundle sheath Rubisco. Establishing an analogous process in a C3 plant would necessitate the introduction of multiple anatomical and biochemical changes (https://c4rice.com/). Substantial progress is being made towards understanding the genetic mechanisms involved in the development of Kranz (C4) anatomy  but introduction of a functional C4 system has yet to be accomplished. If successful, the introduction of a C4 system into hitherto C3 recipients would decrease the amounts of Rubisco—and thus nitrogen fertilizer—required and would also promote water use efficiency . However, running any C4 system requires an additional energy input, and so the resulting crops may be better suited to high light environments.
A simpler approach would be to introduce a CCM that does not require major anatomical changes. The cyanobacterial CCM which concentrates CO2 within Rubisco-containing carboxysomes is relatively straightforward. Introduction of such a CCM into a higher plant would require relatively modest changes to the chloroplasts, namely, the introduction of functional carboxysomes, the removal of stromal carbonic anhydrase and the introduction of HCO3− transporters within the chloroplast envelope. Good progress has been made in the expression of carboxysome-like structures  and of a faster cyanobacterial Rubisco  within higher plant chloroplasts, but further progress is needed to integrate these two processes; to promote HCO3− accumulation in the stroma; and to ensure the synthesis of adequate amounts of functional Rubisco. In parallel with these developments, rapid progress is also being made in identifying and understanding the formation and function of the analogous pyrenoid CCM found in green algae (e.g. ), which offers an alternative approach for engineering more efficient plants .
In C3 plants, the concentration of CO2 in the intercellular spaces is typically 30% lower than in the atmosphere  and the average ratio of chloroplastic to intercellular CO2 is 0.7 . This means that the CO2 concentration in the immediate vicinity of Rubisco is typically 50% of that in the atmosphere. Since even small increases in CO2 concentration will directly increase photosynthetic carbon assimilation, any trait that can increase this is a potential target for crop improvement. CO2 enters the leaves via the stomata, whose conductance is determined by their size, distribution and regulation. Even traits that only indirectly influence stomatal conductance (e.g. root architecture, facilitating access to soil moisture) may promote CO2 assimilation and should be considered . Several different genes that affect stomatal conductance have been identified and successfully manipulated to increase photosynthetic rates . But as CO2 enters leaves, water escapes by transpiration and so improving photosynthetic assimilation by manipulating stomata alone will inevitably increase water consumption, which may be disadvantageous in water-limited environments . However, by increasing the conductance of CO2 from the substomatal spaces to Rubisco, i.e. mesophyll conductance, it should be possible to increase photosynthesis without adversely affecting water use . While the factors involved in mesophyll conductance are not fully characterized, both anatomical (e.g. mesophyll structure, cell wall thickness and chloroplast distribution) and biochemical factors (e.g. patterns of carbonic anhydrase and aquaporin expression) have been implicated . Whereas the anatomical traits are probably the strongest determinants of constitutive mesophyll conductance the biochemical factors are more important in responding to environmental changes .
Free air CO2 enrichment experiments have convincingly shown that increasing photosynthesis does increase crop yields. While photosynthetic carbon fixation is a complex trait, several strategies to increase it have been identified both from in silico and experimental research. However, further targets, for example related to Rubisco biogenesis, regulation and stability, may yet be identified. A number of target traits have already been tested which, using glasshouse conditions, delivered positive results. Under such conditions, it has also been possible to combine several traits for even greater gains. However, the results from a very limited number of field trials using similar material, have been more variable with increases only occurring under certain test conditions (e.g. elevated CO2 or stress). This serves to highlight the need to tailor solutions to specific crops and to the environments in which they will be grown, to deliver the greatest impact. GM approaches will be needed to deliver this, but this will also require overcoming societal concerns over the risks of deploying biotechnology . The relevance of trade-offs between higher carbon fixation capacity and stress tolerance also await evaluation , and may equally depend on wider acceptance of these emerging technologies before being satisfactorily implemented.
Increasing photosynthesis will increase crop yields provided that other constraints are not limiting.
Much of the focus is on directly or indirectly overcoming the limitations of Rubisco.
There is abundant significant variation in Rubisco isolated from different species, and in the properties of Rubisco activase, that could be exploited to increase crop yields.
Increasing the capacity to regenerate RuBP, or decreasing the penalties of photorespiration have worked well in model species.
Ambitious synthetic biology approaches are in progress to create novel CO2 concentrating mechanisms in crop plants.
The authors declare that there are no competing interests associated with the manuscript.
This work was supported by the Biotechnology and Biological Sciences Research Council through the 20:20 Wheat® Institute Strategic Program [grant number BBSRC BB/J/00426X/1; the Lancaster Environment Centre (PhD studentship to G.E.D.); a subcontract to the University of Illinois as part of the award RIPE, Realizing Increased Photosynthetic Efficiency, funded by the Bill and Melinda Gates Foundation (BMGF), the Foundation for Food and Agriculture Research (FFAR) and the U.K. Department for International Development (DFID) (to M.A.J.P. and E.C.-S.).
carbon concentrating mechanism
Rubisco specificity factor for CO2 versus O2