Bringing CAM photosynthesis to the table: Paving the way for resilient and productive agricultural systems in a changing climate

Abstract

Modern agricultural systems are directly threatened by global climate change and the resulting freshwater crisis. A considerable challenge in the coming years will be to develop crops that can cope with the consequences of declining freshwater resources and changing temperatures. One approach to meeting this challenge may lie in our understanding of plant photosynthetic adaptations and water use efficiency. Plants from various taxa have evolved crassulacean acid metabolism (CAM), a water-conserving adaptation of photosynthetic carbon dioxide fixation that enables plants to thrive under semi-arid or seasonally drought-prone conditions. Although past research on CAM has led to a better understanding of the inner workings of plant resilience and adaptation to stress, successful introduction of this pathway into C₃ or C₄ plants has not been reported. The recent revolution in molecular, systems, and synthetic biology, as well as innovations in high-throughput data generation and mining, creates new opportunities to uncover the minimum genetic tool kit required to introduce CAM traits into drought-sensitive crops. Here, we propose four complementary research avenues to uncover this tool kit. First, genomes and computational methods should be used to improve understanding of the nature of variations that drive CAM evolution. Second, single-cell ’omics technologies offer the possibility for in-depth characterization of the mechanisms that trigger environmentally controlled CAM induction. Third, the rapid increase in new ’omics data enables a comprehensive, multimodal exploration of CAM. Finally, the expansion of functional genomics methods is paving the way for integration of CAM into farming systems.

Plants from various taxa have evolved crassulacean acid metabolism (CAM) to deal with water limitation. In this review, we discuss four complementary research avenues for uncovering molecular mechanisms of CAM that may be incorporated into current crops to enhance stress resilience in a changing climate.

Introduction

Economic and environmental sustainability are the foundations of modern agriculture. They are now being challenged by the rapidly growing world population and climate change, with crops suffering from sharply diminishing freshwater resources, rising temperatures, and frequent drought episodes (Arora, 2019; Bogati and Walczak, 2022). These challenges translate into an urgent need to produce more food with fewer inputs (e.g., fresh water) in a sustainable manner. The key to achieving this goal may reside in our understanding of plant photosynthesis and water-use efficiency (WUE), i.e., the amount of biomass or grain that the plant can produce per unit of water used (Briggs and Shantz, 1913).

Crassulacean acid metabolism (CAM) is a well-known example of photosynthetic adaptation to arid conditions. It has been defined primarily by nocturnal gas exchange and C₄ acid accumulation (Haag-Kerwer et al., 1996; Cushman and Borland, 2002; Owen et al., 2016; Winter and Smith, 2022), tight regulation of key CAM genes by the circadian clock (Borland et al., 1999; Lüttge, 2000; Taybi et al., 2002; Dodd et al., 2003; Borland and Taybi, 2004; Boxall et al., 2005, 2017, 2020), and leaf succulence (Koch and Kennedy, 1980; Maxwell et al., 1997; Griffiths et al., 2008; Guan et al., 2020; Heyduk, 2021). Nevertheless, it is noteworthy that this last attribute does not show a definitive positive correlation with CAM in several taxa (Heyduk et al., 2016b; Martin et al., 2019; Leverett et al., 2023).

Unlike C₃ and C₄ plants, CAM plants have evolved the ability to fix carbon dioxide (CO₂) at night. This distinction is made possible by a stomatal opening pattern that is intrinsically linked to the CAM pathway (Hanscom and Ting, 1978; Zeiger et al., 1987; Griffiths, 1989; Borland et al., 2014). The stomata are open for most of the night but remain closed during most of the day. The diel chronology of CAM photosynthesis, which covers a full 24-h continuum, was first described by a four-phase carbon fixation model (Osmond, 1978). Subsequent research has revealed considerable variation in this paradigm, which depends on species, environmental stimuli, and the developmental stage of the photosynthetic organs (Dodd et al., 2002; Osmond et al., 2008). Although the transition between nocturnal and diurnal phases exhibits more flexibility than initially thought, the four-phase model still provides a valuable, albeit generalized, overview of the physiological and biochemical shifts that occur during the CAM cycle. Consequently, it will be used as a generalized framework for the pathway in the following exposition. During phase I, which occurs at night, CO₂ enters the leaves through open stomata (Figure 1)). Phosphoenolpyruvate (PEP) carboxylase (PPC) fixes atmospheric CO₂, resulting in accumulation of oxaloacetate, which is later reduced into malate and stored in the vacuole in the form of malic acid for the remainder of the night. At the end of phase I, the stomata close. Phase II begins at dawn and is characterized by the transition from PPC to Rubisco as the principal carboxylase (Maxwell et al., 1999; Dodd et al., 2002; Matiz et al., 2013). This transition is made possible by simultaneous activation of Rubisco and dephosphorylation of PPC, making the latter sensitive to inhibition by malate efflux from the vacuole (Carter et al., 1990; Jiao and Chollet, 1991; Borland et al., 1999; Nimmo, 2000; Borland and Taybi, 2004). The tight circadian control of phase II ensures the viability of CAM by temporally separating not only the activity of two otherwise competing carboxylases but also competing reactions. Phase III is characterized by the decarboxylation of malate, resulting in the production of CO₂ that can be refixed by Rubisco in the Calvin–Benson–Bassham cycle (Maxwell et al., 1999; Dodd et al., 2002; Borland et al., 2014). Temporal separation of this process from primary carboxylation is essential to avoid wasteful malate cycling. The fourth and final phase begins with the reopening of the stomata, which enables a brief fixation of atmospheric CO₂ through the C₃ pathway. Phase IV also represents the reverse process of the carboxylase shift, with progressive activation of PPC by PPC kinase (PPCK) and inactivation of Rubisco, paving the way for the nocturnal CO₂ carboxylation of phase I.

Figure 1.

Simplified representation of the four CAM phases over a 24-h cycle.

During phase I, which takes place throughout the dark period, stomata open to allow atmospheric CO₂ to enter the cell; CO_{2 is then} hydrated by carbonic anhydrase to produce HCO₃⁻ and fixed by PPC, which is phosphorylated by PPCK. PPC uses PEP and HCO₃⁻ to form OAA, which is rapidly reduced to malate by MDH and stored inside large vacuoles as malic acid for the rest of the dark phase. Phase II represents the transition between phases I and III and occurs between the last hours of the dark phase and the first hours of the light phase. Phase III takes place during the light phase and is characterized by stomatal closure and transition to the C₃ carbon fixation mechanism, as well as inhibition of PPC via dephosphorylation by PP2A and possibly efflux of malate from the vacuole (not shown). In CAM-induced leaves of Mesembryanthemum crystallinum, malate is transported to the cytosol for decarboxylation by NADP-ME during the day. This reaction produces CO₂, which can be fixed by Rubisco in the chloroplast and integrated into the CBB cycle. Pyruvate, the other product of decarboxylation, is converted by PPDK to PEP, which can then be used for gluconeogenesis to replenish the starch pool in the chloroplast. Starch production from triose-P from the CBB cycle is not shown. Phase IV represents the transition from the light to the dark phase. PEP for the subsequent dark phase can be generated by glycolysis of the starch/glucose pool. This model represents the traditional pathway used by starch-forming CAM species and thus illustrates the decarboxylation process carried out by NADP-ME. NAD-ME and the chloroplastic isoform of NADP-ME are not shown. In some plants, malate is oxidized to OAA by MDH and decarboxylated by PEP carboxykinase (PEPCK), leading to direct production of PEP. This figure was partly adapted from Silvera et al. (2010) and produced using Biorender.

Of the various attributes of the CAM cycle outlined above, the reversal in stomatal opening pattern stands out as a crucial factor that contributes to the well-described drought tolerance of CAM plants (Cushman and Borland, 2002; Herrera, 2009; Borland et al., 2014; Fleta-Soriano et al., 2015; Amin et al., 2019). Stomatal closure during the day functions to retain water within cells when evapotranspiration rates would be at their highest in a C₃ plant. As the stomata open at night, when evapotranspiration rates are lower, water loss is minimized (Cushman et al., 2008a; Boxall et al., 2020; Karimi et al., 2021; Winter and Smith, 2022). This characteristic enables CAM plants to achieve significantly higher WUE than their C₃ and C₄ counterparts (Borland et al., 2009, 2014; Silvera et al., 2010). Although this feature is of great interest to agriculture, only a limited number of CAM species have been cultivated as crops. Furthermore, although CAM is a single-cell phenomenon, engineering this mechanism into C₃ or C₄ crops has not yet been reported, as a comprehensive understanding of how the pathway is regulated during each 24-h cycle remains to be achieved.

Obligate, or constitutive, CAM species have been defined as plants in which CAM is always and continuously expressed in mature tissues. By contrast, facultative CAM species have the ability to complete their life cycle using only C₃ or C₄ when grown under favorable conditions, but a switch to CAM photosynthesis can be triggered in response to environmental signals, among which drought stress is the most common (Winter, 2019). The high degree of compatibility between the C₃ and CAM pathways suggests that the changes in facultative CAM species involve remodeling of the biochemistry of individual cells to accommodate a new photosynthetic pathway (Winter, 2019). The genetic elements that regulate this ability may be of considerable value for introduction of CAM into C₃ or C₄ crops. Studying the genomes and epigenomes of facultative CAM species may reveal the mechanisms necessary to trigger this trait.

Seminal ’omics studies of facultative CAM species have provided a wealth of data on the CAM pathway (Kore-eda et al., 2004; Cushman et al., 2008a; Barkla and Vera-Estrella, 2015; Barkla et al., 2016; Brilhaus et al., 2016; Chiang et al., 2016; Barkla et al., 2018; Wai et al., 2019; Ferrari et al., 2020; Kong et al., 2020; Guan et al., 2021). These resources, combined with critical advances in the fields of single-cell ’omics, systems biology, and synthetic biology, create considerable opportunities to generate increasingly comprehensive and well-resolved datasets of CAM plants and/or to delve into unexplored facets of existing datasets. Such studies could not only enable precise study of the full range of molecular mechanisms that lead to CAM induction but also advance our understanding of the evolutionary origins of CAM through comparative genomics. The current state of knowledge presents an ideal opportunity to review the latest research on facultative and obligate CAM species. Here, we propose several avenues of research on CAM species, made possible by the wealth of resources now available to those interested in engineering an inducible CAM system into C₃ and C₄ crops. The proposed methodologies are intentionally described in a non-technical manner, and details of their implementation can be found in the cited literature.

Avenue 1: Phylogeny-informed comparative genomics discovery of CAM innovations

Knowledge of the evolutionary path required for CAM acquisition is critical

CAM evolved from C₃ photosynthesis multiple times and independently across different taxa (Edwards, 2019; Heyduk et al., 2019; Gilman et al., 2023). However, the exact evolutionary path that led to CAM has yet to be unraveled. Two distinct evolutionary models have recently been proposed (Yang et al., 2019). In the first, a single path involves the gradual evolution of C₃ plants into facultative and, ultimately, obligate CAM species. The second model is branched and suggests that obligate and facultative CAM species evolved separately from C₃ species, rejecting the idea that a C₃/CAM intermediate state is necessary for the evolution of an obligate CAM state. A multistep process implies the need to stack distinct (or possibly complementary) mechanisms to achieve facultative CAM and then obligate CAM status. By contrast, a one-step direct transition from C₃ to full CAM may imply the need for fewer molecular changes. Identifying which evolutionary path is the most likely to result in the change from C₃ to CAM will require a detailed phylogeny and trait database. Important biological questions regarding this matter have been raised previously (Edwards, 2019), and their answers must inform future research on CAM. Regardless of which model is correct (and both models may occur in distinct clades of the phylogeny), it will be necessary to collect clear evidence of the influence of plant anatomy on the ability of individual lineages to evolve CAM.

With knowledge of the evolutionary origin, genomes become useful

Knowledge about the evolutionary origins of CAM establishes a framework for rigorous comparative genomics studies. The availability of genomic resources for CAM-performing plants has recently increased with publications of the genomes of the facultative CAM species Mesembryanthemum crystallinum (common ice plant) (Shen et al., 2022) and Sedum album (Wai et al., 2019); the obligate CAM species Ananas comosus (Ming et al., 2015), Phalaenopsis equestris (Cai et al., 2015), Kalanchoë fedtschenkoi (Yang et al., 2017), and Aloe vera (Jaiswal et al., 2021); and the aquatic obligate CAM species Isoetes taiwanensis (Wickell et al., 2021), to name a few. However, these species are generally evolutionarily distant (with the exception of K. fedtschenkoi and S. album, both of which belong to the Crassulaceae family), and more genomes will be needed to fill gaps in the phylogeny and allow for informed comparisons. Most critical will be the availability of genomes of closely phylogenetically related species that differ in their photosynthetic modes. Such relationships have been found within the Yucca genus, which contains similar proportions of C₃ and CAM plants (Heyduk et al., 2016a, 2022). Therefore, species of the Yucca genus provide an opportunity to perform comparative genomic studies on closely related C₃ and CAM species. Additional efforts should be devoted to generating large ’omics datasets for C₃ (such as Yucca filamentosa) and obligate CAM (such as Yucca aloifolia) species from this taxon (Heyduk et al., 2016a, 2019).

With genomes, comparative methods gain power and precision

The expected abundance of genomic datasets will enable identification of the convergent evolutionary signatures implicated in CAM acquisition. Convergent changes have been identified in the PPC2 gene of K. fedtschenkoi (Yang et al., 2017). Although this gene has been shown to have no function in CAM in this species (Boxall et al., 2020), it will be instructive to perform similar studies for genes involved in CAM in a wide range of species that use this pathway. More advanced methods should also be used to characterize the relevance of single signatures and identify all other regions of a genome under the same selective pressure that might be coherent with the evolution of CAM. Because genes subject to selection generally exhibit similar changes in amino acid substitution rates or evolutionary rates, an effective way to identify convergent evolutionary signatures at the genome scale is to use known patterns of rate change. The development of packages such as Relative Evolutionary Rates Converge (RERconverge; Kowalczyk et al., 2019) has made such evolutionary analyses possible. RERconverge tests for associations between rate-change patterns and convergent phenotypes, whether binary or continuous. By filtering candidate regions to retain only those with functions in the CAM pathway, it may be possible to achieve a clear understanding of the nature of the molecular variations that drive the evolution of CAM. The efficacy of additional methods and software in detecting individual convergent sites has been evaluated previously using empirical and simulated data (Rey et al., 2019).

Although generating large quantities of data for CAM plants will be a necessity, it is also essential to use appropriate data-mining methodologies to conduct meaningful comparative studies. The burgeoning field of machine learning (ML) and the new possibilities for its application to plant science should be seized upon by the CAM community. Supervised learning, a type of ML in which data are used to train a model to predict outcomes from large datasets (Soltis et al., 2020), has great potential to answer evolutionary genomics questions (Schrider and Kern, 2018). Such methods have been successfully used to identify regions under selection in the human genome (Schrider and Kern, 2015), supporting the idea that supervised ML could be used in comparative genomics across a broad range of organisms. By providing data on the allele-frequency spectrum (or the genetic variation at given loci) from CAM and C₃ evolutionary pairs as input to train the model, it may be possible to identify genomic regions responsible for CAM plant evolution. Because our annotations and understanding of plant genomes are incomplete, such regions may not be characterized, limiting our ability to draw robust conclusions. Fortunately, a common application of supervised ML in plants involves the prediction of gene functions. This use of ML, recently reviewed in Mahood et al. (2020), promises to make gene annotation a more efficient process. However, CAM is a complex mechanism that will not be fully understood by focusing on only one of its biological aspects.

Another major advantage of using ML in CAM research will therefore be the capability to cover all components of the regulatory and metabolic pathways by integrating genomes with data from proteomic, metabolomic, transcriptomic, epigenomic, and even biogeographic and ecological studies (Xu and Jackson, 2019). One of the main barriers to the application of ML to genomic analysis has been the high price associated with DNA sequencing and generation of large numbers of genomes. Indeed, hundreds of data points, and therefore genomes, will likely be required to develop an efficient ML model, especially for development of supervised learning algorithms. The rapid drop in the price of DNA sequencing over the past decade, as reported by the National Human Genome Research Institute (www.genome.gov/sequencingcostsdata), provides grounds for optimism. Current efforts to develop new sequencing technologies are substantial and suggest that this decline in costs may continue for years to come. This phenomenon is clearly illustrated by the exponential increase in the number of complete plant genomes sequenced each year. Whereas only 26 plant genomes were published between 2000 and 2010, 1031 genomes from 788 plant species were available by the end of 2020, more than half of which were sequenced between 2018 and 2020 (Sun et al., 2022). Should such a scenario occur, the CAM community should seize the opportunity to work together to generate enough CAM species genomes to enable cutting-edge ML in this field.

Nevertheless, although future studies may support the idea that gene gain/loss events led to the evolution from C₃ to CAM, another strong hypothesis could be that differential expression of key regulators led to the repeated evolution of CAM. Given that we still lack evidence to fully support a single theory and that multiple factors may interact to successfully establish CAM, several strategies are worth pursuing (see sections below).

Avenue 2: Discovery of CAM plant transcriptome switches by single-cell genomics

Facultative CAM species for elucidation of transcriptome reprogramming

Approximately 7% of vascular plants are currently estimated to have evolved some level of CAM ability, 3%–4% have evolved to perform C₄, and the remainder use C₃ (Kellogg, 2013; Niechayev et al., 2019; Winter and Smith, 2022; Gilman et al., 2023). Although some commonly observed physiological differences between C₃, C₄, and CAM-performing plants (i.e., the Kranz anatomy of C₄ plants) may raise the question of compatibility between the aforementioned pathways, the fact that C₄ and CAM evolved from C₃ suggests the possibility that every C₃ plant could acquire some degree of C₄ and CAM characteristics (Kellogg, 2013; Edwards, 2019). In addition, the existence of facultative CAM species that can switch from C₃ (Winter and Holtum, 2007; Brilhaus et al., 2016; Holtum et al., 2017, 2018; Heyduk et al., 2019; Winter, 2019; Zhang et al., 2022a; Luján et al., 2022) or C₄ (Moreno-Villena et al., 2022) under certain environmental conditions strongly suggests that it should be possible to engineer the “CAM induction trait” into non-CAM plants. This view is supported by recent evidence that succulent leaf anatomy is not always necessary to accommodate CAM (Heyduk et al., 2021), although it has been proposed that anatomical changes largely determine the potential of a plant to acquire strong CAM biochemistry (Heyduk et al., 2016b; Edwards, 2019). It should be noted, however, that although leaf succulence is unrelated to CAM in some species, CAM might be expected to be less efficient in the absence of leaf and cellular succulence. In addition, the switch to CAM has been shown to be truly reversible in M. crystallinum (Winter and Holtum, 2007; Nosek et al., 2018, 2021), suggesting that crops might be engineered to temporarily induce CAM under water deprivation and revert to their original photosynthetic mode when the stress is no longer present. Finally, the demonstration of switch reversibility implies that plants may not need to acquire or lose specific genes to perform CAM. Instead, rearranging the expression of certain genes or certain regulatory mechanisms (e.g., phosphorylation or dephosphorylation) might be sufficient. Because these complex photosynthetic pathway transitions involve the repurposing of key sets of enzymes and the reorganization of cellular biochemistry, facultative CAM species are ideal systems in which to study how the CAM induction process works. Consequently, one of the proposed objectives for the coming years should be to expand from the foundation (Kore-eda et al., 2004; Cushman et al., 2008a) by applying the latest ’omics techniques and computational methods to achieve a detailed and comprehensive picture of the molecular switches that trigger CAM induction in facultative CAM species.

The potential and necessity of single-cell transcriptomic studies

Traditional “bulk” RNA-sequencing studies can provide only a broad overview of the expression changes in tissues and may conceal the responses of individual cells or cell types. Furthermore, currently available analytical tools for these methods do not allow for accurate monitoring of cell dynamics, as they rely on measuring expression in static cell states. Although some ’omics studies have been performed using sections or peels of M. crystallinum and K. fedtschenkoi leaves that were enriched in specific cell types (Barkla et al., 2016, 2018; Abraham et al., 2020; Guan et al., 2021), this approach does not allow for examination of the cellular changes that occur in response to the environment or the crosstalk between different cell types. These responses and interactions may nonetheless reveal key mechanisms of tolerance to high salinity, such as epidermal bladder cells (EBCs), which are present in aerial tissues of M. crystallinum and play a significant role in Na⁺ and Cl⁻ storage in salt-stressed plants (Barkla et al., 2018). This suggests the existence of a complex transport mechanism between cell types to sequester excess salt in EBCs, and such a mechanism could not be fully studied by isolating EBCs alone. Likewise, the inversion of stomatal movements that occurs during the C₃-to-CAM transition is expected to involve signaling mechanisms and communication between guard cells and other cell types (Wakamatsu et al., 2021). It is therefore essential to perform studies on CAM induction in facultative CAM species at the single-cell resolution. Recent advances in single-cell RNA sequencing (scRNA-seq) technologies and analysis methods have made this possible, as they enable expression profiles to be generated for individual cells at a specific time. By generating scRNA-seq datasets at multiple time points during CAM induction, it would be possible to obtain a dynamic and comprehensive overview of the molecular mechanisms that lead to CAM by focusing the analysis on photosynthetic cell types, and to understand potential adaptations in other cell types.

Single-cell trajectory inference methods as tools to identify key regulators of CAM induction

The advent of scRNA-seq technologies has led to the rapid development of bioinformatic tools to study cellular changes (McKenna and Gagnon, 2019). The highly dynamic nature of the C₃-to-CAM transition makes single-cell trajectory inference (TI) methods perfect tools for tracking the conversion of cells performing C₃ to cells performing CAM (Trapnell et al., 2014; Saelens et al., 2019). These modeling methods rely on the ordering of individual cells by pseudotime, a unit used to measure their progress through a dynamic biological process. This approach could be used to organize cells along a C₃-to-CAM gradient, thereby virtually reconstructing the CAM induction process at single-cell resolution. This approach could be used to reveal key regulators of CAM induction by investigating differential gene expression along this gradient. Such a result may be achieved by merging an scRNA-seq dataset from a CAM species performing C₃ photosynthesis (such as a facultative CAM plant under favorable conditions) with an scRNA-seq dataset from the same species undergoing CAM photosynthesis (the CAM-induced state of the same facultative CAM plant). This approach permits the examination of transcriptional alterations that occur specifically in transitioning cells, minimizing noise from non-transitioning cells and potentially facilitating identification of minimally expressed regulators. The recent development of TI software packages and algorithms that can decipher non-linear developmental trajectories should be seen as an opportunity to study the process of CAM induction at unprecedented temporal and spatial resolution. Among the numerous pipelines now available, the most widely used TI methods were recently evaluated and compared in a comprehensive review (Saelens et al., 2019). Although Slingshot (Street et al., 2018) had the highest overall performance for tree trajectory types, partition-based graph abstraction (Wolf et al., 2019) worked best for connected and disconnected graphs and SCORPIUS for linear trajectories (Cannoodt et al., 2016; Saelens et al., 2019). By uncovering transcriptome changes at critical points of the pseudotime trajectories, the results of such analyses can not only provide crucial information about the key elements that drive cells to engage in the transition but also reveal some of the genetic components involved in the evolution from C₃ to CAM photosynthesis (Heyduk et al., 2019). Similarly, the recent development of single-cell trajectory tree alignment tools (Alpert et al., 2018; Conde et al., 2022; Sugihara et al., 2022) opens the door to a wide range of applications in research on facultative CAM. Comparing trajectory trees from closely related C₃, obligate CAM, and facultative CAM species could provide answers to major biological questions about photosynthetic adaptation to drought. Although TI analysis could fulfill its promise of providing keys to answering some of the most pressing questions in developmental biology, there are some caveats to this technique. To compute a trajectory, the user must choose a starting or ending point in the form of a cell or cluster. It is also possible to input both the beginning and end cells of the trajectory and identify the potential lineages that connect them (Wu and Zhang, 2020). Because of this biased approach, the discussed techniques cannot provide information about the direction of the trajectory and should be used only to reconstruct developmental processes that have already been well studied and characterized. RNA velocity analyses applied to scRNA-seq datasets can predict the future state of individual cells based on a pre-mRNA/mRNA ratio (La Manno et al., 2018). Estimates of RNA velocity can thus help anticipate the direction of trajectories by providing indications of the transcriptional direction of each cell in a two-dimensional space. This method can therefore be used to further refine previous TI analyses. Applying this tool to studies of developmental trajectories leading to EBC formation in M. crystallinum could lead to substantial progress in research on salt tolerance (Winter, 2019). A five-step pipeline for identifying the main regulators of CAM induction using scRNA-seq is presented in Figure 2.

Figure 2.

Discovery of CAM-plant transcriptome switches by single-cell genomics.

Proposed process for analyzing expression changes during the transition from C₃ to CAM in facultative CAM species. A prolonged period without water triggers the induction of CAM in some plant species. Isolation of cells or nuclei from leaves at key points in the transition for single-cell RNA sequencing (1) and integration of the resulting datasets enable the visualization of distinct C₃ and CAM cell clusters based on the differential expression of CAM genes (3). Trajectory analyses of the transition in gene expression between cells performing C₃ and those performing CAM photosynthesis can be used to identify regulators of the induction process (4). Comparison of trajectories with those observed in datasets of closely related obligate C₃ and CAM species may help validate candidate regulators and identify additional ones (5). Figure produced using Biorender.

Using these methods to study not only M. crystallinum but also obligate CAM species could greatly enhance our understanding of the pathway. For example, certain species of the genus Kalanchoë, whose leaves show a gradual developmental progression from C₃ to CAM, could prove to be highly informative systems (Hartwell et al., 2016). Indeed, ’omics studies of the environmentally triggered transition in facultative CAM plants will certainly be more difficult to interpret, as CAM genes can be expressed alongside a number of stress-response genes. Furthermore, plants such as Guzmania monostachia, Agave deserti, Agave tequilana, and potentially Agave sisalana have been reported to exhibit a longitudinal gradient of CAM expression along the length of a leaf, meaning that these species may also be cleaner systems for studying CAM induction (Freschi et al., 2010; Dubinsky, 2013; Gross et al., 2013; Yin et al., 2018). Such gradients have the potential to be captured and clearly represented using state-of-the-art spatial transcriptomics methods. These technologies enable quantification of transcripts at the precise locations where they are expressed within tissue sections (Burgess, 2019; Rao et al., 2021; Tian et al., 2023). The Visium spatial transcriptomics platform from 10× Genomics, for example, was recently used to resolve the functioning of C₄ + CAM photosynthetic metabolism in Portulaca oleracea (Moreno-Villena et al., 2022).

Because candidate transcription factors (Heyduk et al., 2018; Yin et al., 2018; Amin et al., 2019) and ion homeostasis–related transporters (Tran et al., 2020; Li et al., 2021) are predicted to play critical roles in CAM regulation, future studies on expression changes should place more emphasis on identifying and characterizing genes with related functions at a single-cell resolution. Such findings would provide essential information for pinpointing the exact nature of the signaling pathways and transcriptional controls of CAM induction. A clear understanding of cell-type-specific responses to drought would promote development of informed strategies for bioengineering of CAM traits into crops and minimize pleiotropic effects on plant growth, development, and productivity.

Avenue 3: The potential of new proteomics, metabolomics, and epigenomics techniques

From chromatin structure to regulation: Connecting genome accessibility to gene expression patterns

Several studies of DNA methylation in CAM plants suggest the strong involvement of epigenetic mechanisms in regulation of CAM photosynthesis (Duarte-Aké et al., 2016; Shi et al., 2021) and CAM induction in M. crystallinum (Dyachenko et al., 2006; Huang et al., 2010). Only a few studies have attempted to uncover the role of chromatin accessibility in CAM plants, although the epigenetic state of chromatin has repeatedly been shown to influence many plant biochemical and physiological processes (Alvarez et al., 2019; Thiebaut et al., 2019; Farmer et al., 2021; Tian et al., 2021). To fully elucidate a process such as CAM induction, it is therefore essential to study the epigenome. Furthermore, chromatin structure studies have proven to be effective in identifying cis-regulatory elements and transcription factors (Lu et al., 2018; Farmer et al., 2021), which may play significant roles in triggering a photosynthetic switch to CAM (Zhang et al., 2022b). The introduction of single-cell assay for transposase-accessible chromatin using sequencing (scATAC-seq) methods (Buenrostro et al., 2015; Lake et al., 2018; Lareau et al., 2019), as well as software packages to analyze the resulting data (Fang et al., 2021), will pave the way for another promising avenue of CAM research and provide an opportunity to integrate data from different single-cell ’omics studies (e.g., histone post-translational modification [PTM] proteomics).

From proteins to pathways: Using proteomics and phosphoproteomics to unravel the complexity of CAM induction

The past decade has seen the emergence of quantitative proteomic and metabolomic studies of M. crystallinum at the single-cell-type level, most of which have focused on EBCs (Barkla and Vera-Estrella, 2015; Barkla et al., 2016, 2018). In addition, a proteomics study of Agave leaves across the CAM diel cycle and a comparison between the epidermis and the mesophyll leaf proteomes of K. fedtschenkoi have recently been published (Abraham et al., 2016, 2020). A thorough understanding of proteomic networks is essential, as plant responses to abiotic stresses typically involve processes such as protein–protein interactions and PTMs (Kosová et al., 2018; Chaudhary et al., 2019). The need to map the proteomes and phosphoproteomes of additional CAM-performing plants is reinforced by the well-documented fact that reversible phosphorylation of PPC is a central process of the CAM pathway (Nimmo et al., 1987; Carter et al., 1991; Boxall et al., 2017). Furthermore, pyruvate orthophosphate dikinase (PPDK) activity is reported to be inhibited at night through phosphorylation by PPDK-regulatory protein (RP1), illustrating the importance of PTMs in diurnal regulation of CAM (Dever et al., 2015). Efforts in this area could be accelerated by the rapidly growing potential of mass spectrometry (MS)-based quantitative proteomics and phosphoproteomics methods (Taylor et al., 2021; Perron et al., 2022). In the past few years, important progress has been made in MS imaging (Van Acker et al., 2019), phosphopeptide enrichment (Abe et al., 2017; Yi et al., 2018; Ahmed et al., 2019; Zheng and Jia, 2019; Finamore et al., 2020), and data analysis (Beekhof et al., 2019; Ressa et al., 2019; Bouwmeester et al., 2020; Deznabi et al., 2020; Locard-Paulet et al., 2020; Savage and Zhang, 2020). Drafts of the Arabidopsis proteome and phosphoproteome were recently produced using immobilized metal-ion affinity chromatography followed by liquid chromatography–tandem MS and uncovered 8577 phosphoproteins, representing 47% of the proteome (Mergner et al., 2020). These results clearly underline the importance of studying protein phosphorylation in plants. In addition, rapid advances in the field of single-cell proteomics (Saha-Shah et al., 2019; Kelly, 2020; Petelski et al., 2021; Brunner et al., 2022) should be followed closely by the plant community. The recent development of automated MS imaging methods for mapping protein expression in plant cells at a high spatial resolution brings single-cell proteomic studies in sight and provides an important opportunity to complement our understanding of the CAM pathway (Liang et al., 2018; Taylor et al., 2021). A recent application of MS-based proteomics techniques in M. crystallinum highlighted the differential protein changes that occur in guard cells and mesophyll cells during CAM induction, emphasizing that each cell type appears to undergo independent induction events, as illustrated by differences in protein expression (Guan et al., 2021). The current revolution in proteomics offers an opportunity to rapidly decipher the inner workings of drought tolerance in CAM plants by identifying the critical phosphorylation events that lead to CAM induction or by deciphering the regulation of CAM proteins by the circadian clock. Such information, coupled with single-cell transcriptomic data, would be valuable for identifying critical regulatory mechanisms of CAM.

From metabolites to mechanisms: Advancing CAM research with mass spectrometry-based metabolomics

The accumulation of various metabolites such as pinitol, ononitol, myo-inositol, and proline to maintain osmotic balance in response to certain abiotic stresses is well characterized in M. crystallinum, and the study of this process has recently regained importance (Vernon and Bohnert, 1992; Barkla and Vera-Estrella, 2015; Tran et al., 2020; Ceusters et al., 2021; Zahedi et al., 2021). Rapid MS advances have made it possible to perform metabolomic studies at a single-cell resolution. The development of optical-fiber-based laser ablation electrospray ionization MS enables the creation of comprehensive metabolic profiles of individual cells in situ (Stopka et al., 2018). Additional advances in single-cell metabolomics and their potential application to studies of plant stress adaptation have recently been reviewed by Katam et al. (2022). Examining the metabolic diversity of individual plant cells subjected to different environmental conditions may provide additional information on the mechanisms used by CAM plants to withstand drought and heat and should be another focus in the coming years.

Avenue 4: Synthetic biology and genome editing advances provide a stepping stone for CAM engineering in crops

Functional genomics analyses are essential for understanding CAM evolution and complementing ’omics studies

As mentioned above, a thorough understanding of the primary evolutionary trajectories that have led to CAM is necessary to accurately inform subsequent studies of this pathway. However, the execution of such an endeavor is contingent upon establishing the functional roles of each molecular component. Furthermore, although the hypotheses that emerge from extensive ’omics datasets hold the potential to catalyze significant advances in the field, these analyses will become relevant only when complemented by functional validation studies in planta. A persistent obstacle to the completion of this objective is the limited availability of routinely transformable and regenerable CAM species (Hartwell et al., 2016). Owing to their relative ease of cultivation, stable transformation, and propagation, several Kalanchoë species such as K. fedtschenkoi, K. daigremontiana, and K. laxiflora currently serve as model CAM species (Borland et al., 2009; Hartwell et al., 2016) and have provided the basis for a number of functional studies of the CAM pathway (Hartwell et al., 1999, 2002; Dever et al., 2015; Boxall et al., 2020). In addition, M. crystallinum, which serves as a model for facultative CAM, was the target of the first mutagenesis screen of a CAM-performing species, which culminated in the identification of CAM-deficient mutants lacking plastidic phosphoglucomutase (Bohnert et al., 1988; Bohnert and Cushman, 2000; Cushman et al., 2008b). Although additional transformable CAM model species would undeniably increase the precision of functional studies, it is essential for the community to harness currently available transformation methods to validate putative regulators predicted by ’omics studies. Indeed, the accumulation of vast datasets without functional validation is undesirable, as it would raise innumerable untested, and sometimes competing, hypotheses about potential regulators of CAM genes, making it challenging to pursue relevant information.

Utilizing CAM mechanisms to unlock the potential of crops

The holy grail of CAM forward engineering, or CAM biodesign, is to equip C₃ and C₄ crops with characteristics of the CAM pathway to improve their WUE and resilience in the face of prolonged drought (Yuan et al., 2020). However, scientists attempting to reach this ultimate objective will need to answer major evolutionary questions about the innovations involved in CAM acquisition before starting their quest. Does successful engineering of CAM in drought-sensitive crops require introduction of CAM genes in their original form? Alternatively, could it be sufficient to rewire existing networks by modifying the expression profiles of existing genes or PTMs of proteins? Although the rapid and common evolution of this trait may suggest that simple modulations in the expression of one or a few genes may be sufficient, addressing these questions is crucial to establishing effective engineering strategies. Specifically, the pursuit of distinct goals, such as substituting specific amino acids or altering expression and PTMs, requires different methods and tools. Successful CAM biodesign will first require an understanding of how CAM genes are activated in the right cells at the right time, how functional protein complexes are created, and how regulation of these complexes is carried out at the right time in the 24-h CAM cycle. To achieve this goal, the community will need to translate computational ’omics data into biological data by delving into the realm of systems biology/functional genomics.

Uncertainty about the evolutionary origin of CAM has been illustrated by the emergence of two contrasting approaches to CAM engineering in recent years. In the first approach, key genes of the CAM machinery are transformed into plants that perform C₃ or C₄. This approach has recently been put into practice through Agrobacterium transformation of Arabidopsis thaliana with selected M. crystallinum CAM genes from the carboxylation (i.e., beta-carbonic anhydrase [BCA2], PPC1, PPCK1, NAD-dependent malate dehydrogenase [NAD-MDH1, NAD-MDH2], and NADP-dependent malate dehydrogenase [NADP-MDH1]) and decarboxylation modules (i.e., NAD- and NADP-dependent malic enzymes [NAD-ME1, NAD-ME2, NADP-ME1, NADP-ME], PPDK, PPDK-RP, and PEP carboxykinase [PEPCK]; Lim et al., 2019). Each of these candidates was expressed individually in Arabidopsis. This study should therefore not be regarded as a genuine attempt at CAM engineering of a C₃ plant but as a guideline providing important information on the nature of certain components that will need to be considered in future attempts at CAM biodesign. As expected, overexpression of individual CAM genes from ice plant carboxylation and decarboxylation modules in Arabidopsis did not yield evidence of CAM-related phenotypic changes. This is because the introduction of genes encoding not only CAM enzymes but also those involved in carbohydrate metabolism, transport and storage, and maintenance of cytoplasmic homeostasis, as well as appropriate transcription factors and other regulators, is required for optimal and timely expression of the pathway (Borland et al., 2014; Winter and Smith, 2022). By contrast, the second approach involves rewiring genes already present in C₃ and C₄ crops to enable induction of CAM upon exposure to drought or salinity stress, as many (or all) CAM-regulating genes may already be present in plants of all taxa (Yin et al., 2018; Heyduk et al., 2019; Liu et al., 2020). Nevertheless, the presumption that all CAM-related genes are ubiquitously present across plant species is grounded in data derived from incomplete genome annotations. Numerous genes encoding proteins of currently undetermined functions are induced concomitantly with CAM, and only a small number of these genes have been subjected to thorough investigation. Future detailed characterization of such genes may reveal elements unique to CAM.

The multitude of technologies emerging from current revolutions in synthetic biology and genome editing may not only enable new attempts to reprogram C₃ and C₄ crops to perform CAM but also contribute significantly to characterization of key CAM genes across species. In addition, single-cell transcriptomic and scATAC-seq data from CAM-performing plants can be used to obtain valuable insights into the most appropriate genetic engineering strategies. Such data allow scientists to target the expression of specific genes using stress-inducible promoters in specific cell types, avoiding many of the negative pleiotropic effects observed when using generic or constitutive promoters.

Genome editing techniques for large-scale functional characterization of CAM genes

The ability of CRISPR–Cas9 methods to generate knockin and/or knockout mutants for functional genomics studies was recently examined in the CAM model species K. fedtschenkoi (Liu et al., 2019). Such studies will assuredly multiply in the coming years, providing essential information to the community. Other applications of CRISPR–Cas9 may involve reprogramming essential genes directly involved in the CAM pathway, as well as transcription factors (Cushman and Bohnert, 1992; Amin et al., 2019) and microRNAs (Chiang et al., 2016; Wai et al., 2017), in C₃ and C₄ crops to make them capable of autonomously inducing CAM under certain conditions. As demonstrated by its recent use to silence PPC1 in K. laxiflora (Boxall et al., 2020), RNA interference is another useful tool for assessing the different degrees of involvement of CAM genes in different species. The recent publication of M. crystallinum transformation protocols (Hwang et al., 2019; Agarie et al., 2020), coupled with new genome-editing and synthetic biology techniques, can help accelerate the functional characterization of genes essential for enhancing stress resilience and WUE in major crops.

Concluding remarks

Although the primary intent of this perspective was not to provide a comprehensive synthesis of historical research on CAM photosynthesis, it aimed to propose strategies for future research on facultative and obligate CAM plants. Furthermore, the bulk of this article focused on state-of-the-art systems biology techniques. Thus, the anatomical and physiological requirements of CAM were not discussed in as much detail as the molecular and biochemical aspects of the CAM system. The development and rapid adoption of single-cell ’omics, systems biology, and synthetic biology approaches and the development of highly efficient comparative ’omics analysis tools may propel CAM species to the forefront of plant science. The recent publication of genomes from facultative and obligate CAM plants opens the door to implementation of new analytical applications and should encourage efforts to increase the genomic resource pool of CAM species. As the identification of facultative CAM plants has accelerated in the past decades, research should be extended to these newly characterized species. For example, certain species in the genus Clusia are the only trees known to perform CAM photosynthesis, and the flexibility of CAM induction in species such as Clusia minor L. raises many questions about the mechanisms underlying this special capability (Lüttge, 2008; Luján et al., 2022; Pachon et al., 2022). Further research on CAM induction in woody dicot plants may provide a unique opportunity to engineer this trait in trees of significant interest for biomass production, such as poplar. Furthermore, deciphering the inner workings of the switch to CAM in plants of different taxa will undeniably contribute to completing our understanding of CAM induction. The rapid increase in applications of artificial intelligence to biology should also be seen as an opportunity to unravel the complexity of this process. It is not possible to limit our view of CAM to a single aspect of the pathway. Instead, it is imperative to confront expression data with proteomic, metabolomic, epigenomic, and phenotypic data. The potential of artificial intelligence to integrate data from different modalities has recently been documented (Lipkova et al., 2022; Qiao et al., 2022) and promises to help the CAM research community make great strides in modeling and prediction of photosynthetic adaptations. Major themes, such as the number of genes that must be modified to enable induction of CAM in C₃ and C₄ crops or the manner in which CAM evolves from C₃, will be addressed to guide subsequent meaningful transformation studies. Likewise, routine stable transformation methods must be developed for different CAM plants to further accelerate the identification and characterization of important CAM genes.

Funding

This work was supported by the University of Florida CALS Dean’s Award to N.P. and the new faculty startup fund of the University of Mississippi to S.C.

Published: November 21, 2023