Engineering crassulacean acid metabolism to improve water-use efficiency

Abstract

Climatic extremes threaten agricultural sustainability worldwide. One approach to increase plant water-use efficiency is to introduce crassulacean acid metabolism (CAM) into C₃ crops. Such a task requires comprehensive systems-level understanding of the enzymatic and regulatory pathways underpinning this temporal CO₂ pump. Here, we review the progress that has been made in achieving this goal. Given that CAM arose through multiple independent evolutionary origins, comparative transcriptomics and genomics of taxonomically diverse CAM species are being used to define the genetic ‘parts list’ required to operate the core CAM functional modules of nocturnal carboxylation, daytime decarboxylation, and inverse stomatal regulation. Engineered CAM offers the potential to sustain plant productivity for food, feed, fiber, and biofuel production in hotter and drier climates.

Photosynthesis for a parched planet

Earth’s population is projected to exceed 9 billion by 2050. The consequent demands on agriculture for food, feed, fiber, and fuels, coupled with decreasing arable land area and increasing nitrogen and phosphate fertilizer requirements for crop production all point to the need to produce more plant-derived biomass with reduced resource inputs [1]. Approximately 40% of the world’s land area is considered arid, semi-arid, or dry sub-humid, with precipitation amounts that are inadequate for most conventional agriculturally important C₃ or C₄ crops [2]. Although water is the most crucial resource for sustainable agriculture, projections regarding global warming suggest that there could be a gradual increase in the severity and frequency of extreme weather conditions, including higher temperatures and drought conditions [3-5]. Prolonged drought and overreliance on groundwater for crop irrigation has led to the depletion of aquifers in the USA [6, 7] and in other regions of the world [8, 9].

Multiple strategies have been proposed to improve agricultural productivity via enhancing photosynthesis [1, 10, 11]. These include, but are not limited to, the introduction of: (i) either single-cell or two-celled C₄ photosynthesis into C₃ plants [12-14]; (ii) a CO₂-transporting aquaporin [15]; (iii) inorganic carbon-concentrating mechanisms (CCMs) from cyanobacteria [16, 17]; (iv) a CCM shared by all eukaryotic photosynthetic organisms for active recycling of photorespiratory CO₂ from mitochondria to chloroplasts [18]; (v) synthetic carbon fixation pathways [1, 19, 20]; and (vi) strategies to reduce photorespiration during carbon fixation [21]. By contrast, relatively little attention has been paid to the potential of moving crassulacean acid metabolism (CAM) into C₃ plants. CAM is a temporally controlled plant inorganic CCM that maximizes water-use efficiency (WUE) by shifting all or part of the CO₂ uptake to the nighttime, when evapotranspiration rates are reduced compared with the daytime. This review assesses the progress that has been made in defining the genetic requirements and strategies for the assembly and operation of CAM in C₃ plants via synthetic biology.

CAM – a strategic target for synthetic biology

The major plant inorganic CCMs in terrestrial vascular plants are CAM and C₄ photosynthesis [22]. CAM arose through multiple, independent evolutionary origins in at least 343 genera across 36 plant families representing >6% of higher plant species [23]. CAM resembles C₄ photosynthesis in its use of C₄ organic acids as storage intermediates during carbon fixation, but exploits a temporal separation of primary and secondary CO₂ fixation. Furthermore, CAM maximizes WUE by concentrating CO₂ around ribulose-1-5-bisphosphate carboxylase/oxygenase (RUBISCO), favoring carboxylase activity (Figure 1). The two most distinctive features of CAM are: (i) nocturnal CO₂ uptake and fixation by phosphoenolpyruvate carboxylase (PEPC) in the cytosol, which leads to the formation of C₄ organic acids that are stored in the vacuole and (ii) an inverse stomatal behavior, in which stomata are closed during part of or all of the day and are open at night. The organic acids accumulated overnight are subsequently decarboxylated during the day to release CO₂, which is refixed by RUBISCO in the chloroplast, leading to carbohydrate production via the C₃ Calvin–Benson cycle (Figure 1).

Figure 1.

A simplified view of crassulacean acid metabolism (CAM). Stomata open at night allowing atmospheric CO₂ to enter the cell to be captured as bicarbonate (HCO₃⁻) by cytosolic phosphoenolpyruvate carboxylase (PEPC) leading to the formation of C₄ acid, which undergoes protonation and is stored in the vacuole as malic acid. Stomata are closed for all or part of the subsequent day owing to malic acid efflux from the vacuole and decarboxylation by the malic enzyme to release CO₂, which is then refixed by plastidic ribulose-1,5- bisphosphate carboxylase/oxygenase (RUBISCO) through the Calvin–Benson photosynthetic carbon reduction cycle. C₃ acids produced in the cytosol serve as substrates for the gluconeogenic pathway during which carbohydrates are regenerated and may be stored as starch (shown) or other storage carbohydrates.

For convenience, the CAM day-night cycle is defined classically as four separate phases of gas exchange, the onset or duration of which can be reduced or eliminated based upon prevailing environmental conditions [24, 25] or upon the degree of tissue succulence as shown by system dynamics modeling [26]. Phase I is the period of high stomatal conductance and nocturnal CO₂ uptake and assimilation of atmospheric and respiratory CO₂ by PEPC into oxaloacetate (OAA). OAA is subsequently reduced to malate by NAD(P)-malate dehydrogenase and stored in the vacuole as malic acid. Under natural conditions, this phase results in improved WUE as the leaf-air vapor pressure deficit (VPD) is usually lower than during the day [25]. The rate of nocturnal CO₂ uptake is limited by mesophyll processes, such as carboxylation capacity derived from storage carbohydrates [27, 28] or vacuolar storage capacity, instead of by stomatal conductance [29]. Phase II is defined as the transition from dark C₄ carboxylation by PEPC to daytime RUBISCO-mediated carboxylation as photosynthetically active radiation (PAR) increases in the early morning [25]. During phase II, a combination of CO₂ from the atmosphere and that released from organic acid decarboxylation is fixed. Phase III is the period of daytime C₃ photosynthesis when net CO₂ uptake falls to zero [24] and of major decarboxylation of C₄ acids by either NAD(P)-malic enzyme or PEP carboxykinase depending on the species [29]. CO₂ liberated during this phase is concentrated behind closed stomata from 2- to 60-fold in the vicinity of RUBISCO [30], essentially creating a ‘CO₂ pump’, which can potentially reduce photorespiration, a process that can decrease photosynthesis by up to 40% in C₃ plants [31]. The 3-carbon compounds (e.g., pyruvate or PEP) released by malate decarboxylation are converted into storage carbohydrates, which reach peak accumulation at the end of the light period. Phase IV occurs towards the end of the daylight period when partial net CO₂ uptake recommences as a result of the depletion of nocturnally accumulated organic acids, and leads to a decline in leaf internal partial pressure of CO₂ (p_i) and an increase in stomatal conductance and transpiration [25]. As in phase II, both C₃ and C₄ carboxylation reactions can occur although C₄ carboxylation increases as the dark period nears [25]. In summary, CAM can dramatically limit water loss and enhance the magnitude and duration of net CO₂ uptake over a 24-h cycle in resource-limited environments [32]. Requirements for optimal CAM operation are summarized in Box 1.

Box 1. Requirements for CAM.

• Nocturnal CO₂ uptake: depending on the species and environmental conditions, all, most, or the majority of CO₂ uptake by CAM plants occurs during the nighttime when leaf-air VPD is low [25].

• Inverse day–night pattern of stomatal closure: to accommodate nocturnal CO₂ uptake, stomata open in the dark (phase I) and for variable periods of time at the start (phase II) and end (phase IV) of the photoperiod, depending on plant water status and species, but close during the daytime (phase III) [25].

• Diel accumulation–depletion of organic acids: the primary carboxylation reactions result in the nocturnal accumulation of C₄ organic acids (mainly malate), which are subsequently degraded to provide an internal CO₂ source during the subsequent photoperiod [28].

• Reciprocal turnover of storage carbohydrates: storage carbohydrates, such as starch, glucans, soluble hexoses and/or disaccharides, exhibit accumulation patterns that are reciprocal to the pattern of C₄ organic acid accumulation [28]. Up to 20% of leaf dry weight can be committed to carbohydrates for supplying phosphoenolpyruvate (PEP) to CAM [136].

• Enhanced expression of C₄ anabolism enzymes: carbonic anhydrase (CA), which converts CO₂ to HCO₃⁻; the primary carboxylation enzyme phosphoenolpyruvate carboxylase (PEPC), which converts HCO₃⁻ to oxaloacetate (OAA); and malate dehydrogenase (NADP⁺- and NAD⁺-MDH), which converts OAA to malate, each exhibit increased mRNA and protein expression and enzyme activities relative to those in C₃-performing plants [58, 76].

• Enhanced expression of C₄ catabolism enzymes: the primary decarboxylation enzymes NADP⁺-/NAD⁺-malic enzyme (ME), which convert malate to pyruvate with the release of CO₂ in the cytosol with subsequent conversion of pyruvate to PEP by pyruvate orthophosphate dikinase (PPDK), or decarboxylation of malate to PEP by phosphoenolpyruvate carboxykinase (PEPCK) depending on the species, each exhibit increased mRNA and protein expression and enzyme activities relative to those in C₃-performing plants [57, 58, 76].

• Enhanced expression of glycolytic–gluconeogenic pathway enzymes: coordinated expression of a suite of glycolysis and gluconeogenesis pathway enzymes to supply substrates for nocturnal primary carboxylation and daytime decarboxylation (e.g., PEP, pyruvate) reactions, respectively, is essential for CAM [76, 136-138].

• Transport activities between subcellular organelles: the operation of the metabolic sequence of CAM requires the orchestration of transporters at the tonoplast (e.g., voltage-gated inward-rectifying malate channel, tonoplast dicarboxylate transporter, vacuolar H⁺-ATPase), mitochondrial, and chloroplast envelope membranes [83, 139, 140].

• Leaf or stem succulence: some degree of leaf succulence, characterized by increased mesophyll cell size owing to large storage vacuoles and increased mesophyll tissue and leaf thickness to ensure a high capacity for nocturnal organic acid storage [107]. Large cell size, tightly packed cells, reduced intercellular air spaces, and reduced surface area exposure to these air spaces are likely to limit CO₂ diffusion within these tissues.

• Circadian clock control of CO₂ fixation: CAM plants show robust rhythms of CO₂ uptake that persist under constant light and temperature conditions, indicating that these rhythms are under circadian clock control. Expression of mRNAs and post-translational regulatory events, such as the reversible phosphorylation of PEPC by PEPC kinase, are also clock controlled [59, 81, 137].

The inherently high WUE of CAM plants signifies their potential for sustainable production of biomass in a warmer and drier world [33, 34]. Highly succulent CAM species, such as Agave spp. or Opuntia ficus-indica, have been grown commercially for centuries as sources of fiber, sugars for alcohol-containing beverages, and as food or animal forage and fodder in semi-arid and arid regions of the world. More recently, the recognition of the ability of these species to operate at near-maximum productivity with relatively low requirements for water [33, 34] and nutrient inputs has engendered scientific interest in their use as sustainable bioenergy feedstocks [35-39]. These CAM crops avoid competition for existing land resources because they can be grown on marginal or degraded land with poor soil conditions, where precipitation totals or frequency are insufficient to support traditional C₃ or C₄ crops [35-37, 40]. Furthermore, CAM plants could be used for sustainable production on irrigated lands using up to 80% less water to produce similar amounts of biomass compared with C₃ species [32, 33]. Thus, research into expanding the agricultural uses of CAM species should be a high priority to ensure that adequate food, feed, and fiber needs are met in future warmer climates with diminishing arable land and water resources.

Engineering of CAM

The development of bioenergy feedstocks and food crops engineered with the improved WUE of CAM plants complements the direct use of CAM species to supply human needs. CAM and C₄ photosynthesis have been described as products of either parallel or convergent evolution of a complex trait with the implication that many, if not all, of the genes and some regulatory elements necessary for these photosynthetic specializations are already present in C₃ species [23, 41-43]. Such reasoning also underpins the ambitious aims and rationale for transferring C₄ properties to C₃ plants as a means of enhancing plant productivity to achieve food security [12, 14]. Whereas optimal performance of C₄ photosynthesis requires specialized anatomy, including bundle sheath and mesophyll cells to accommodate the spatially separated reactions of C₃ and C₄ carboxylation [41], CAM might prove more tractable because it is a single-cell adaptation requiring only mesophyll cells in contrast to two-cell C₄ photosynthesis. Many permutations of CAM have been described within the evolutionary continuum of CAM species including CAM idling, CAM cycling, weak CAM, latent CAM, facultative or inducible CAM, and obligate or constitutive CAM [23] . In facultative CAM species, CAM expression is readily modified by salinity, water deficit, and high light [29]. Furthermore, some CAM species are capable of switching from C₃ to CAM and back to C₃ [34], which implies that there are no metabolic incompatibilities between C₃ photosynthesis and the water-conserving adaptation. Thus, there appears to be no a priori reason why the pathway cannot be engineered into non-CAM crops as a means of enhancing water-use efficiency or increasing carbon balance.

A logical goal for engineering CAM is the installation of a complete, obligate CAM pathway as this would likely maximize WUE. However, intermediate steps in CAM engineering might also be considered beneficial. Some CAM-cycling or facultative CAM species are thought to benefit from a partial commitment to CAM, not by increasing net carbon gain, but by simply maintaining a positive carbon balance by reducing respiratory CO₂ losses. Such variants of CAM increase WUE and water absorption, resulting in an extension of the plant’s life cycle and thus improve reproductive success under water-deficit stress [44]. Thus, strategies to incrementally engineer a partial CAM pathway might provide partial benefit to C₃ plants. For example, expression of an engineered Solanum tuberosum PEPC under the control of a dark-induced promoter from Arabidopsis thaliana resulted in Arabidopsis plants with greater stomatal conductance, respiration, and transpiration in dark-adapted leaves, and increased CO₂ assimilation rates under different external CO₂ concentrations (C_a) and light intensities compared to wild-type plants [45].

Probing phylogenetically diverse lineages to enable comparative CAM genomics

A fundamental requirement for engineered CAM is to first understand the minimal set of genes and proteins required for its efficient establishment and operation. Until recently, there has been a paucity of genome or transcriptome information available for CAM species. However, the genomic sequences and transcriptome atlases available from cycling, facultative, or obligate CAM species sampled from diverse phylogenetic origins should rapidly redefine our understanding of the molecular genetics of CAM (Figure 2).

Figure 2.

Phylogenetic relationships among major model and agronomically important CAM species for which genomic-scale transcriptomic or genomic sequence datasets are available. Phylogeny derived from the Angiosperm Phylogeny Website (http://www.mobot.org/MOBOT/research/APweb/). Phalaenopsis equestris image © 2008 AltonX. Ananas comosus image © 2009 Enviromantic. Opuntia ficus-indica image © 2005 martinalfaro. Sedum album image © 2007 Frank Vincentz.

Among monocots, the partial transcriptomes or genomes of CAM orchids in the genus Phalaenopsis have been characterized [46-50]. The transcriptome [51] and genome (R. Ming, personal communication) of pineapple (Ananas comosus) are also being sequenced. RNA-sequencing (RNA-seq) and quantitative whole-transcriptome analysis has been performed in young (C₃) and mature (CAM) leaves of Agave americana (X. Yang, unpublished) and across the leaf developmental gradient in several Agave species, including Agave deserti [52], Agave sisalana [53] (J. Hartwell, unpublished), and Agave tequilana [52, 54]. Systematic analysis of gene expression patterns along such developmental gradients are expected to reveal putative regulatory factors involved in the establishment and daily optimization of CAM, in a manner similar to the establishment of the developmental profile of C₄ photosynthesis in maize (Zea mays) leaves [55, 56].

Among core eudicots, several species have emerged as genetic and ecophysiological models for CAM, including the common ice plant (Mesembryanthemum crystallinum), in which CAM is induced following the imposition of salinity or water-deficit stress and Clusia minor, in which CAM is rapidly inducible and reversible [57]. Comparative messenger RNA (mRNA) expression profiling experiments in M. crystallinum performing C₃ and CAM have revealed changes in the abundance of mRNAs encoding key enzymes for CAM [58]. However, the use of facultative CAM species requires that mRNA responses to the stresses used to induce CAM (e.g., high salinity or water deficit) must be distinguished from those associated specifically with CAM.

Within the Crassulaceae, obligate CAM species, such as Kalanchoë fedtschenkoi, Kalanchoë daigremontiana, and Kalanchoë laxiflora, develop CAM as leaves expand from the shoot apical meristem [59], even under well-watered conditions [60]. Quantitative RNA-seq comparisons are underway between developmentally immature, C₃-performing leaves and mature, CAM-performing leaves in these species to discover CAM-specific genes (J. Hartwell, unpublished). Kalanchoë species are readily transformable [61, 62], and thus represent a powerful model system for probing the function of genes and regulatory elements essential for CAM via transgenic RNA interference and overexpression approaches. The genome and transcriptome sequence data being generated for K. fedtschenkoi and K. laxiflora should dramatically expand the possibilities for CAM functional genomics within Kalanchoë. Genome sequence information has been generated for another member of the Crassulaceae, the inducible CAM-cycling species Sedum album [63] (T.P. Michael, unpublished). Lastly, partial transcriptome information is available for a second member of the Caryophyllales, Opuntia ficus-indica [64], which is the most widely cultivated member of the cactus family [65].

Comparative CAM genomics

Comparative transcriptomic and genomic approaches can be used to discern CAM gene function by comparing the expression patterns of known CAM enzymes and transporters among closely related C₃ and CAM species within the same genus or family that show C₃, weak CAM, and strong CAM [23]. Comparative genome and transcriptome sequencing studies of C₃ and C₄ model species from diverse taxonomic origins do not support the duplication or expansion of C₄ pathway genes or the long-standing hypothesis that C₄ photosynthesis has been facilitated by gene duplications followed by neofunctionalization [42]. Similarly, comparative analysis among CAM species is expected to provide informative examples of either parallel or convergent evolutionary events leading to CAM as transcriptome and genome sequence information becomes available (Figure 2).

Another strategy for resolving CAM gene function is to study DNA polymorphisms associated with CAM phenotypes among diverse genotypes arising from either natural or artificial populations within the same CAM species. Ideally, a reference genome sequence should be available so that re-sequencing of hundreds of individual genotypes is feasible and affordable. Embracing this strategy, the research community studying CAM is working with he DOE Joint Genome Institute to establish K. laxiflora as an Arabidopsis-like CAM model species. K. laxiflora has several traits that make it useful as a model, including that it: (i) is diploid (n = 17) with a relatively small genome (~250 Mb), (ii) has relatively small stature (30 to 45 cm tall), (iii) has a short life cycle (~ 6 months), (iv) is easily transformed, and (v) is self-compatible, with the ability to produce many thousands of small seeds per plant. Although M. crystallinum has been studied extensively as a facultative CAM model, its genome is larger (~390 Mb) [66] than that of K. laxiflora and M. crystallinum is not readily transformed.

Coexpression network modeling of CAM

A key challenge for the engineering of CAM into C₃ plant species lies in understanding the temporal regulatory events controlling not only the core carboxylation–decarboxylation of C₄ acids, but also the coincident metabolic fluxes through glycolysis–gluconeogenesis, storage carbohydrate synthesis and breakdown, as well as stomatal control, over the course of the day-night cycle as illustrated in Figure 3A. A multi-layer approach incorporating transcriptional data, functional genomics annotation, and genetics within an integrative modeling framework might afford the best means to discover genes comprising a functional CAM module (i.e., a set of gene or gene products related by their participation in a common biological process) and then to translate this information into well-informed biodesign strategies.

Figure 3.

Temporal dynamics of the major physiological and biochemical processes and key mRNA expression patterns of CAM. (A) Gas exchange, which occurs mostly at night resulting in the peak accumulation of C₄ acids at dawn, is measured as titratable acidity, and reciprocal accumulation of starch or soluble sugars, which peak at dusk [28]. (B) Carboxylation module showing the accumulation of steady-state mRNA for chloroplastic (and possibly cytosolic) β-carbonic anhydrase (CA), cytosolic phosphoenolpyruvate carboxylase (PEPC), cytosolic PEPC kinase (PPCK), and multiple NAD(P) malate dehydrogenase [NAD(P)-MDH] genes, which show similar mRNA expression patterns in M. crystallinum and thus are represented together [58, 81]. (C) Decarboxylation module showing the accumulation of steady-state mRNA for cytosolic NADP-dependent malic enzyme (cNADP-ME), chloroplastic NADP-dependent malic enzyme (cpNADP-ME), mitochondrial NAD-dependent malic enzyme (mtNAD-ME), and plastidic pyruvate orthophosphate dikinase (PPDK). Expression data summarized from CAM-performing M. crystallinum [58, 81] with the exception of mtNAD-ME (J.C. Cushman, unpublished).

Putative functional gene groups and their associated regulation can be inferred by combining multiple layers of functional genomics data into gene network models [67, 68]. Network models are constructed as simple node-edge graphs in which genes and proteins are connected by an edge if they are deemed to be associated or similar to one another in some manner using experimental data. Network gene modules are loosely defined as densely connected structures in a node-edge graph. Gene coexpression networks created solely from relative mRNA abundance data can declare a gene pair or cohort to be associated based on user-defined similarity metrics and expression thresholds. This process results in the construction of coexpression networks of genes that are closely connected through potential co-regulation or functional association.

Protein–protein interaction networks offer other approaches to identify genes with related functions or shared regulation [68, 69]. Such networks are built upon interactions that are either known experimentally [70] or predicted by data mining methods that detect protein-protein interologs, which are interacting pairs of homologous proteins conserved across different species [69, 71, 72]. Integration of protein–protein interaction and coexpression network modules with Gene Ontology [68] or transcription factor regulatory interaction data [69] are approaches well-suited for extracting gene modules of similar function. The integration of functional genomics data, particularly gene coexpression network data with phylogenetic information, referred to as ‘systems genetics’, is an emerging discipline aimed at understanding the molecular mechanisms governing complex traits [73]. Although promising, the integration of functional genomic data alone does not provide evolutionary insights from phylogenomics data. The establishment of complete genomes and transcriptomes for phylogenetically ordered CAM species (Figure 2), along with numerous genetic resources for various C₃ species, now make systems genetics a viable strategy for CAM biodesign. Lastly, integration of phenotypic or genotypic data with gene networks using machine-learning algorithms has been shown to improve the identification of genetic elements contributing to mouse weight [74] and human disease-causing alleles [75]. With increasing genetic and genomic resources for CAM species, exploration of such approaches is expected to be fruitful in the context of biodesign efforts.

Biodesign of CAM modules

Although network-modeling approaches can provide information about new candidate genes by virtue of their association with known genes within a functional module, empirical testing of minimal functional modules, coupled with information from loss-of-function studies of individual enzymes, regulatory proteins, or transcription factors can provide important empirical information about the basic genetic requirements for CAM biodesign. For simplicity, a set of discrete functional modules for carboxylation and decarboxylation, and stomatal control, as well as anatomical requirements for CAM, can be designed and tested as a set of minimal functional modules rather than on a gene-by-gene basis to accelerate the empirical testing process. The rationale for engineering a complete module at one time is that single-enzyme engineering is unlikely to result in a functional CAM pathway with the desired improvements in WUE. This is illustrated by attempts to engineer constitutive, nocturnal expression of PEPC in Arabidopsis, which resulted in only incremental improvements in CO₂ assimilation rates compared to wild type plants [45].

Nocturnal carboxylation module

Nocturnal CO₂ uptake via open stomata and primary fixation as HCO₃⁻ in the dark (phase I) requires the coordinated action of a set of CAM-specific enzymes, some of which have multiple subcellular locations, along with a regulatory kinase [i.e., phosphoenolpyruvate carboxylase (PEPC) kinase (PPCK)] (Figures 3A,B and 4). In the facultative CAM model M. crystallinum, the induction of CAM results in a 2- to 30-fold increase in the transcript abundance of CAM-specific enzymes depending on the particular isogene in question [58]. Corresponding enzyme activities have also been shown to increase relative to those in C₃-performing plants [76]. However, evidence for rhythmic patterns in the abundance of corresponding proteins (e.g. PEPC and RUBISCO) over the light–dark cycle is lacking [77, 78]. Although RUBISCO is activated by RUBISCO activase (RCA) in response to light and circadian clock signals [79], PEPC is activated as a result of allosteric regulation modulated via N-terminal phosphorylation of the enzyme [59]. In CAM plants, the circadian clock activates PEPC kinase (PPCK) transcription and translation in the dark, leading to enhanced phosphorylation of PEPC. Phospho-PEPC has reduced sensitivity to feedback inhibition by malate, and this phosphorylation helps to sustain CO₂ fixation for the majority of the dark period [59, 80, 81]. Thus, engineering an efficient CAM carboxylation module will require the introduction of a clock-controlled, dark-phased PPCK, which should help to prevent a futile cycle of malate synthesis and decarboxylation. The temporal phasing of requisite mRNA expression patterns is illustrated in Figure 3B. It is generally thought that CAM-specific enzymes would possess the kinetic properties required for CAM, which may not be present in C₃ non-photosynthetic isoforms, as has been described in C₄ plants [82]. Correct temporal expression in the target host C₃ species will require cis-regulatory expression patterns with circadian clock control to drive expression of these enzymes only during the dark period.

Figure 4.

Proposed model of a minimum gene set encoding enzymes of core CAM carboxylation–decarboxylation modules and their known regulation based upon enzyme activity [76, 141, 142] and mRNA expression studies [58, 59, 81]. Key metabolites include: G6P, glucose-6-phosphate; PEP, phosphoenolpyruvate; OAA, oxaolacetate; MAL, malate; PYR, pyruvate. Broken arrows indicate multiple metabolic steps. Key enzymes include: CA, carbonic anhydrase; PEPC, phosphoenolpyruvate carboxylase; PPCK, PEPC kinase; NAD(P)-MDH, NAD or NADP-dependent malate dehydrogenase; NAD(P)-ME, NAD- or NADP-dependent malic enzyme; PPDK, pyruvate orthophosphate dikinase; PPDK-RP, PPDK regulatory protein. The involvement of PPDK-RP is inferred from studies in C₃ and C₄ species [91]. Colors indicate subcellular locations of enzymes: cytosol (orange), chloroplasts (green), and mitochondria (blue).

Another key consideration for a fully functional carboxylation module is the possible need to engineer enhanced malate transport into the vacuole during phase I, if insufficient nocturnal vacuolar malate accumulation is observed following testing of the core carboxylation module (Figure 4). Malate influx is presumably mediated by one or more voltage-gated inward rectifying malate channels [83] each with functional redundancy and discrete regulatory responses for controlling malate influx and efflux [84]. Such channels have been characterized in Arabidopsis mesophyll (AtALMT9) [85] and guard cells (AtALMT6) [84]. Furthermore, malate transport in mesophyll vacuoles of Arabidopsis is accomplished by a tonoplast dicarboxylate transporter AttDT [86, 87]. However, the genes or gene products encoding these channels and transporters have not been described functionally for any CAM species to date. Therefore, an important goal of ongoing transcriptomic and genomic sequencing efforts is the molecular identification of these malate channels and transporters.

Daytime decarboxylation module

Daytime CO₂ release from decarboxylation of stored malate and refixation via the Calvin cycle (phase III) also requires the coordinated action of a set of CAM-specific enzymes. The release of CO₂ from malate can be catalyzed by one of several NAD(P)-malic enzymes (MEs), some of which exhibit elevated and circadian clock-controlled mRNA expression patterns (Figure 3C) and distinct subcellular locations (Figure 4). In M. crystallinum, these CAM-specific enzymes show up to 8-fold increases in mRNA abundance [58] depending on the isogene, along with corresponding increases in enzyme activities relative to C₃-performing plants [76]. In CAM plants, pyruvate orthophosphate dikinase (PPDK) is required to recycle pyruvate from malate decarboxylation to PEP. PPDK can be localized to the chloroplast, to the cytosol, or to both compartments depending on the genus or species [88, 89]. PPDK mRNA and protein expression are increased during C₃-to-CAM induction in M. crystallinum [58, 76, 90]. PPDK-regulatory protein (RP) is a bifunctional kinase/phosphatase that catalyzes the reversible phosphorylation–dephosphorylation of PPDK over the light–dark cycle, leading to inactivation–activation of PPDK in C₃ and C₄ plants [91].

As in the design of the carboxylation module, a key consideration for a fully functional decarboxylation module is the possible need to engineer enhanced malate efflux from the vacuole during phase III (Figure 4). However, the process of malic acid efflux from the vacuole is not well understood. Efflux might occur by passive diffusion, by a proton-linked symporter, or by a tonoplast dicarboxylate transporter (AttDT) [86, 87] potentially regulated by reversible phosphorylation events [83]. Therefore, the molecular identification and functional characterization of putative vacuolar efflux transporters is needed.

Stomatal regulation module

In CAM plants, the inverse day–night pattern of stomatal closure and opening that underpins the high WUE of the pathway has been linked to the substantial changes in the leaf p_i generated during the diel process of malate turnover [92, 93]. At night, stomata are thought to open in response to the drawdown in leaf p_i as PEPC is activated. Furthermore, the expression of a PEPC engineered to exhibit reduced malate sensitivity indicated that both constitutive and dark-induced expression of this ‘CAM-like’ PEPC in Arabidopsis resulted in enhanced stomatal opening and transpiration rates at night [45]. Such efforts illustrate the need for engineering coordinated activation–deactivation of carboxylases–decarboxylases over the diel cycle to ensure CAM-like stomatal regulation. During the day in the CAM leaf when stomata close, up to 10,000 μmol CO₂ mol⁻¹ may be generated via malate decarboxylation [92]. Stomatal closure is thought to be driven by a signaling cascade wherein CO₂ is sensed either directly by guard cells, by leaf mesophyll cells, or both cell types [94].

Defining the cell-type specificity of CO₂ signaling components will require more research. In brief, guard cell-specific β-CARBONIC ANHYDRASE (βCA1, βCA4) and the HIGH LEAF TEMPERATURE 1 (HT1) kinase mediate early stomatal closure signaling events that converge with downstream ABA- and Ca²⁺-signaling transduction networks that act on SLOW ANION CHANNEL ASSOCIATED 1 (SLAC1) S-type (and R-type) anion channels to trigger anion efflux from the guard cells. This anion efflux results in membrane depolarization, which drives K⁺ efflux from guard cells via outward-rectifying K⁺ _out channels, resulting in stomatal closure [94]. The idea that changes in the concentration of malate in the apoplast of C₃ plants might mediate guard cell responses to [CO₂] by activating H ⁺-out efflux channels and anion channels in guard cells and thereby facilitate the efflux of the osmoregulatory anions Cl⁻ and malate²⁻ during stomatal closure [95, 96] was first proposed in 1993-1994 [97, 98]. More recent evidence from C₃ plants with genetically altered flux through the TCA cycle supports the concept of mesophyll-derived, apoplastic [malate] as a reporter of leaf p_i, and thus, a key effector for linking mesophyll and stomatal function [99, 100].

The diel variation in stomatal responsiveness to leaf p_i that has been reported for some Kalanchoë species [101] implies that circadian gating of guard cell responsiveness to apoplastic [malate] might be required to provide a further layer of control over CAM stomata. An important task that needs to be undertaken as part of CAM biodesign will be to identify the genes and proteins that regulate the transport of malate between the cytosol and the vacuole, and between the cytosol and the apoplast, and to establish the appropriate level of circadian control over these potential checkpoints for linking stomatal regulation with mesophyll metabolism.

The ‘malate as CO₂ sensor’ idea supports the concept that the mesophyll regulates guard cell function [102], and that it should thus be possible to bioengineer the CAM-defining nocturnal opening and daytime closure of stomata in C₃ plants by installing the enzymatic machinery responsible for diel turnover of malate in the leaf mesophyll. However, alterations in C₃ guard cell metabolism or signaling processes might also be required.

Possible divergence of signaling networks between C₃ and CAM guard cells is suggested by reports of the insensitivity of CAM stomata to blue light, inferred from studies with individual leaves or epidermal peels from the facultative CAM species Portulacaria afra and M. crystallinum [103-105]. Defining the guard cell transcriptome, proteome, and metabolome in a CAM model species might reveal whether the signaling or metabolic hubs and networks of C₃ guard cells need to be re-engineered to achieve the inverse day–night pattern of stomatal conductance that typifies CAM.

Anatomical requirements for CAM

Leaf or stem succulence, a concomitant anatomical trait with CAM, results from the presence of large cell vacuoles that are presumed to be a prerequisite for the storage of malic acid during the night [106, 107]. Within the relatively undifferentiated leaves that typify most CAM species, increased succulence tends to reduce internal air space and the surface area of chloroplast-containing mesophyll cells directly exposed to intercellular air spaces [107, 108]. These anatomical traits reduce leaf internal conductance to CO₂ and direct uptake of atmospheric CO₂, but enhance carbon economy during decarboxylation in phase III of CAM because net CO₂ efflux from the leaf is minimized [109]. At first glance, this ‘trade-off’ between the optimal leaf anatomy for CAM and the ideal internal structure for C₃ photosynthesis would appear to present a challenge for engineering CAM in C₃ crops without incurring significant penalties in yield potential. However, a study of functional leaf anatomy within the genus Clusia, which includes C₃, C₃–CAM, and constitutive CAM species, indicates that distinct layers of palisade and spongy mesophyll present further options for accommodating both direct uptake of CO₂ via RUBISCO and CAM [34]. The relatively well-aerated spongy mesophyll of Clusia helps to optimize direct C₃-mediated CO₂ fixation, whereas the enlarged and densely packed palisade cells accommodate the potential for C₄ carboxylation and nocturnal storage of organic acids [34]. These findings indicate that selecting genotypes of C₃ target crops with increased ploidy level, which is correlated expediently with increased cell size and biomass productivity [110], succulence [66], and a well-developed palisade mesophyll, in principle, should expedite engineering of CAM into C₃ crops. However, if such target genotypes are unavailable, the mesophyll cells could be selectively enlarged by introducing basic helix–loop–helix transcription factors that increase cell size [111]. Alternatively, expression of a putative xyloglucan endotransglucosylase hydrolase gene could result in an increase in leaf water storage and succulence, alongside an increase in the number of mesophyll cells and reduced intracellular air space without incurring an increase in leaf thickness [112].

Identifying target host species

Initial CAM biodesign efforts will target the genetic model Arabidopsis, or close relatives, owing to its rapid growth rate and ease of transformation. With regard to bioenergy feedstocks, preferred targets are rapid-cycling C₃ crops, such as members of the Brassicaceae, particularly oilseed crops and fast-growing woody plants within the Populus genus, which are used extensively in the timber, pulp, and paper industries, and more recently as a bioenergy crop. There are many genetic and genomic resources available for Populus [113], including several sequenced genomes [114, 115] and well-developed transformation systems [116, 117]. There is also substantial variability in leaf anatomy and morphology among Populus spp. [118, 119]. Moreover, leaf cell size in Populus is stimulated by free-air CO₂ enrichment [120]. These results indicate that Populus leaf cell size has the potential to be increased through genetic modification strategies, as outlined above. Improvement of the WUE of Populus by the introduction of CAM would enable sustainable agroforestry production and possibly expansion of production acreage into more arid regions.

Moving complex traits into target host species

Many biological processes are controlled by gene modules composed of an array of genes [121, 122]. Similarly, complex trait engineering, such as the proposed CAM engineering project outlined here, will require the use of multigene stacking technologies. Most plant genetic engineering attempts have been limited to the introduction of one or a few genes at a time [123]. To address this limitation, new methods have been developed recently to assemble multigene plant transformation vectors that include a zinc-finger nuclease and homing endonuclease [123], in vivo site-specific assembly [124], recombination-assisted multifunctional DNA assembly [125], or a standardized assembly system based on type IIS restriction enzymes that allows the indefinite expansion of reusable gene modules made from standardized DNA components [126, 127]. The multigene plant transformation vector approach has one major drawback: the maximum number of genes in each vector is limited by the cloning capacity of the recipient vectors. Although the TAC vector, derived from the bacteriophage P1 cloning system, is capable of accepting DNA fragments of up to 100 kb [128] and the BIBAC vector, derived from the bacterial artificial chromosome system, is capable of maintaining a genomic DNA fragment of 150–300 kb [129, 130], the assembly of such large constructs with a large number of genes remains a challenge [131]. Multigene plant transformation vector approaches have successfully transferred fewer than 10 genes at a time [123, 124]. However, in planta gene stacking by site-specific recombination has advantages for multigene transfer because of its ability to effectively resolve complex transgenes into precise, single-copy insertions at known genomic target sites [132]. In planta site-specific recombination is limited to stacking one gene at a time and would be time-consuming to use for stacking more than 10 genes. Mini-chromosomes (or artificial chromosomes) provide another vehicle for stacking multiple genes and offer several advantages. Transgenes are grouped into a closely linked block, which avoids linkage drag [131]. Both ‘top-down’ (i.e., involving engineering of the existing chromosomes within a cell) and ‘bottom-up’ approaches (i.e., de novo assembly of chromosomes from centromeric arrays, telomere repeats, and replication origins) can be used to create mini-chromosomes [131, 133]. Although mini-chromosomes have been constructed for genetic transformation in maize [134], Arabidopsis [135], and rice (Oryza sativa) [131], this new technology is not yet ready for routine gene stacking in Populus [133]. Efforts are being made to develop mini-chromosome technology in Populus, and the centromeric regions of several chromosomes have been identified in the Populus genome assembly (X. Yang, unpublished). Implementation of such large-scale cloning and multigene stacking technologies will ensure that each CAM module will be assembled and expressed efficiently within host genomes.

Conclusions and future directions

The ability of succulent CAM species, such as Agave and Opuntia, to maintain high biomass productivities with water input of only 20% of that required by C₃ or C₄ crops has drawn attention to their possible use as feedstocks for biofuel production. Indeed, such CAM species can be grown in areas where precipitation is typically insufficient to support C₃ or C₄ crops. Thus, exploring the agricultural uses of CAM species should bear fruit as global warming continues to erode finite arable land and water resources.

The transfer of the WUE of the CAM to C₃ crops represents an exciting alternative to the direct use of CAM crops, thereby creating the theoretical potential for sustainable agricultural production on semi-arid lands and avoiding competition with current food and biofuel production systems. One limitation for engineering CAM is the lack of a comprehensive ‘parts list’ required for the design of functional nocturnal carboxylation, daytime decarboxylation, and inverse stomatal regulation modules. However, the rapid expansion of comparative ‘omics’ resources across a phylogenetically diverse set of monocot and eudicot CAM species is imminent. Comparative genomic analysis of this wealth of information is expected to aid in the identification of evolutionarily conserved enzymes and regulatory factors that constitute the minimal requirements for a functional CAM system. Such information will in turn aid the design and transfer of CAM to C₃ crops.

Key challenges for successfully engineering a functional CAM system into C₃ crops in the near future will include improved understanding of: (i) temporal or circadian clock-controlled regulation of enzyme activities besides those known for PEPC and PPDK in the carboxylation and decarboxylation modules, respectively; (ii) the functional characterization and temporal control of malate channels and transporters that mediate the nighttime influx and daytime efflux of malate into and out of the vacuole; (iii) the signaling pathways that control nighttime opening and daytime closing of stomatal guard cells; (iv) the temporal regulation of the glycolytic and gluconeogenic pathways and associated transporter activities, which feed C₃ substrates to carboxylation and from the decarboxylation modules, respectively; (v) the regulation of the hydrolytic and phosphorolytic pathways of starch degradation and starch biosynthesis enzyme activities, which supply the glycolytic and gluconeogenic pathways, respectively; (vi) in soluble sugar-accumulating CAM species, regulation of the activity of tonoplast sugar transporters, which move sugars into and out of the vacuole; (vii) RNAi-mediated transcriptional and posttranscriptional regulatory circuits, if present; and (viii) posttranscriptional and posttranslational regulation of the synthesis and turnover of protein components of the CAM machinery. CAM module design and implementation will also need to be coupled with leaf tissue anatomy and succulence to obtain a fine balance between CO₂ diffusion into the leaf and CO₂ trapping and concentration within the leaf to optimize the efficient operation of CAM. Lastly, cloning technologies for the efficient assembly or stacking and transfer of multigene constructs into target host species are improving rapidly, and are expected to enable the successful engineering of CAM into Arabidopsis and Populus, a highly productive bioenergy feedstock, as well as other food and fiber species in the near future.

Highlights.

• Crassulacean acid metabolism (CAM) species exhibit high water-use efficiency (WUE).

• Introduction of CAM into crops should serve to enhance water-use efficiency and increase carbon balance.

• Genome and transcriptome sequencing of diverse taxa of CAM species is in progress.

• Coexpression network modelling can help define the genetic requirements for CAM biodesign.

• Carboxylation, decarboxylation, and stomatal regulation module engineering will be required.