Comparative genomics can provide new insights into the evolutionary mechanisms and gene function in CAM plants.

Crassulacean acid metabolism (CAM) photosynthesis is an important biological innovation enabling plant adaptation to hot and dry environments. CAM plants feature high water-use efficiency, with potential for sustainable crop production under water-limited conditions. A deep understanding of CAM-related gene function and molecular evolution of CAM plants is critical for exploiting the potential of engineering CAM into C3 crops to enhance crop production on semi-arid or marginal agricultural lands. With the newly emerging genomics resources for multiple CAM species, progress has been made in comparative genomics studies on the molecular basis and subsequently on the evolution of CAM. Here, recent advances in CAM comparative genomics research in constitutive and facultative CAM plants are reviewed, with a focus on the analyses of DNA/protein sequences and gene expression to provide new insights into the path and driving force of CAM evolution and to identify candidate genes involved in CAM-related biological processes. Potential applications of new computational and experimental technologies (e.g. CRISPR/Cas-mediated genome-editing technology) to the comparative and evolutionary genomics research on CAM plants are offered.

Introduction

Crassulacean acid metabolism (CAM) is a CO2-concentrating mechanism in plants as an adaptation to hot and dry environments. CAM plants can be divided into two main categories: (i) facultative CAM plants, which exclusively or predominantly perform C3 photosynthesis (or C4 photosynthesis in some instances) under well-watered conditions, with CAM induced or up-regulated in a reversible manner in response to reduced water availability caused by drought or high salinity stress, and (ii) constitutive CAM plants, in which CAM is always expressed in mature photosynthetic tissues (Winter, 2019). In comparison with other types of photosynthesis, CAM provides a unique mechanism resulting in high water-use efficiency (WUE), which can be 3–20-fold and 2–10-fold that of C3 and C4 plants, respectively (Borland ). The high WUE of CAM plants results from special temporal patterns of CO2 fixation and stomatal movement with four temporally separated phases. Phase I is the period of opening of stomata for primary atmospheric CO2 fixation via phosphoenolpyruvate carboxylase (PEPC) at night when temperature is lower. Phase II takes places at the transition from dark to dawn, featuring a shift from C4 carboxylation by PEPC to ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO)-mediated CO2 assimilation. Phase III is the period of RuBisCO refixation of CO2 released from malate decarboxylation, with stomata closing during the daytime when temperature is higher. Phase IV occurs toward the end of the light period when stomata re-open to allow net CO2 uptake as a result of the exhausting of malate reserves (Osmond, 1978). Furthermore, during the daytime CAM concentrates CO2 released from malate in the vicinity of RuBisCO and consequently reduces photorespiration mediated by RuBisCO, a process that can decrease photosynthesis by up to 40% in C3 plants (Ehleringer and Monson, 1993), resulting in a theoretical biochemical efficiency of CAM higher than that of C3 photosynthesis (Davis ). Inspired by the potential for enhancing the sustainable production of food and biomass on semi-arid, abandoned, or marginal agricultural land (Borland , 2014; Yang ), rich genomics resources have been created for multiple CAM species, including Phalaenopsis equestris (Cai ), pineapple (Ming ), Agave (Gross ; Abraham ), Kalanchoë (Yang ), Erycina (Heyduk ), Yucca (Heyduk ), Talinum (Brilhaus ), Portulaca (Christin ), and Sedum (Wai ). Leveraging these genomics resources, the CAM research community has made progress in comparative genomics research determining gene function originating in CAM plants and investigating the molecular signatures of CAM evolution. In this review, we summarize the recent progress in comparative analysis of DNA/protein sequences and diel gene expression patterns to identify candidate genes involved in CAM-related biological processes (e.g. CO2 fixation, stomatal movement, circadian rhythm). Also, based on the new insights gained from linking genes to the differences between C3 and CAM physiology, we update our understanding of the molecular mechanisms underlying the evolution of CAM from C3 plants.

The course of CAM evolution

It is widely accepted that CAM has evolved from C3 photosynthesis (West-Eberhard ). However, the path of C3-to-CAM evolution remains unclear. From the metabolic perspective, it was recently argued that C3-to-CAM evolution represents a continuum and the requirements for CAM (e.g. nocturnal CO2 uptake, enzyme expression for CO2 fixation, storage of organic acids, associated transport activities, carbon flow through glycolytic and gluconeogenic pathways) are all present in C3 plants and only need enhancement for transition from C3 to CAM (Bräutigam ). This argument oversimplifies the complexity of the molecular mechanisms underlying the C3-to-CAM transition. Recently, a genome-wide comparative genomics analysis of three constitutive CAM plant species (Ananas comosus, Kalanchoë fedtschenkoi, and P. equestris) in comparison with non-CAM species predicted that changes in at least 60 genes (i.e. rewiring of diel gene expression pattern in 54 genes and protein sequence mutations in another six genes) were potentially involved in CAM evolution (Yang ). In comparison with C3 species, changes in the expression of a large number of protein-encoding genes were also revealed in other comparative analyses of constitutive CAM plant species such as Agave (Abraham ; Yin ) and Ananas comosus (Ming ). Additionally, many genes, including transcription factors, are potentially involved in CAM induction in facultative CAM plants such as Mesembryanthemum crystallinum (Amin ). Besides protein-encoding genes, microRNAs (miRNAs) are likely involved in CAM gene regulation (Wai ). Although analysis of phenotypical diversity suggests the hypothesis that there is a C3–CAM continuum, the genotypic evidence does not support this hypothesis (Winter, 2019). Therefore, we propose two alternative models for CAM evolution: (i) a single-path model, in which the evolution of constitutive CAM was mediated by facultative CAM (Fig. 1A), or (ii) a multi-path model, in which facultative CAM and constitutive CAM evolved independently from C3 (Fig. 1B). The path of C3-to-CAM evolution is still controversial. Even though there are plant families and even single genera (e.g. Clusia) containing facultative CAM, weak CAM and strong CAM species, we cannot exclude the possibility that facultative CAM and constitutive CAM evolved independently from a common C3 ancestor. Further experimental research needs to be conducted to test the hypotheses in Fig. 1 by addressing the following questions: Can any C3 species be directly converted to facultative CAM, weak CAM, or strong CAM? Can any facultative CAM species be directly converted to weak CAM or strong CAM? Can any weak CAM species be directly converted to strong CAM? If yes, how many and what genes will need to be engineered through overexpression, knockout/knockdown, and/or rewired temporal expression?

Fig. 1.

A conceptual framework for crassulacean acid metabolism (CAM) evolution and working hypotheses for guiding comparative genomics research. (A) Hypothesis (i): CAM evolution followed a linear course leading from facultative CAM to strong constitutive CAM. (B) Hypothesis (ii): facultative and constitutive CAM evolved independently.

It can be assumed that facultative CAM and weak CAM have a major difference in the regulation of gene expression, with the expression of CAM-related genes induced by water-deficit condition in facultative CAM plants whereas the expression of CAM-related genes is mainly under ontogenetic control in weak CAM plants. The conversion of facultative CAM to weak CAM would involve a major change in the regulatory mechanism, which is not trivial from the perspective of genetic engineering.

Besides the aforementioned changes in protein sequence and temporal/spatial gene expression required for C3-to-CAM evolution, it was recently hypothesized that anatomical modifications (e.g. an increase in leaf succulence) are the rate-limiting step in the evolution of strong CAM whereas the evolution of facultative CAM or weak CAM does not require major anatomical changes (Edwards, 2019; Heyduk ). It would be necessary to identify the genes responsible for the anatomical modifications required for strong CAM evolution.

The driving force of CAM evolution

The interplay of historical contingency (i.e. the outcomes of processes are unpredictable and sensitive to random events) and natural selection (i.e. a deterministic process reflecting systematic differences in the propensity of alternative genotypes to survive and reproduce, depending on their fitness in a given environment) is the driving force of evolution (Blount ). Although historical contingency and natural selection have been demonstrated to be responsible for the outcomes of evolution in various organisms, their relative contribution to the rich diversity of life on our planet is still a hot topic of debate. This is true even within the context of CAM evolution. CAM has evolved from C3 photosynthesis many times independently from diverse plant lineages (Edwards and Ogburn, 2012), featuring phenotypical convergence in physiological and metabolic attributes, e.g. PEPC-mediated CO2 fixation at night and the inverted day/night pattern of stomatal movement (Yang ). Since CAM occurs within various phylogenetic hierarchies, including entire orders (Edwards and Ogburn, 2012), entire families (Simpson, 2010), or entire genera (Crayn ), we hypothesize that historical contingency, as defined above, has played a role in CAM evolution. This hypothesis is further supported by the following genomic information: (i) in comparison with non-CAM species, single-site convergent mutations were found in only two of the three distantly related CAM species studied by Yang (A. comosus, K. fedtschenkoi, and P. equestris), and (ii) no significant expansion of gene families is shared among these three CAM species (data not shown). However, in comparison with a C3 species (Arabidopsis), convergent change in the diel expression patterns of more than 50 genes was found in two distantly related CAM species studied (K. fedtschenkoi and A. comosus) (Yang ), suggesting that natural selection on gene expression has contributed to the regulatory mechanisms enabling the C3-to-CAM conversion. Caution should be taken in interpreting the results from this comparative analysis of limited number of species because it is possible that the number of genes showing convergent change in the diel expression patterns may be reduced when more plant species are added to the comparative analysis in the future. Recently, through comparative analysis of orthologous gene pairs between CAM species Agave americana and three non-CAM species, including two C3 species (Arabidopsis and Oryza sativa) and one C4 species (Zea mays), Yin identified 64 genes that were shown to carry at least one positively selected site in the protein sequences, including phosphoenolpyruvate carboxylase kinase 1 (PPCK1), which plays an important role in PEPC-mediated CO2 fixation in CAM plants. Therefore, both protein sequence mutations and changes in diel gene expression patterns were under natural selection during CAM evolution. According to this information, we can assume that both historical contingency and natural selection are important factors driving CAM evolution.

Gene duplication (at the levels of the individual gene, chromosomal segment, or entire genome) is a major force for generation of evolutionary novelties leading to adaptation to the environment (Kanazawa ). Although no recent genome-wide duplication has been detected in the A. comosus genome (Ming ), recent whole-genome-duplication events were found in several other CAM lineages, including Agave (Ming ) and K. fedtschenkoi (Yang ). With such massive gene duplication events functional diversification is expected to follow (Yang ). In K. fedtschenkoi, diversification in diel expression profiles of duplicated genes were revealed in multiple CAM-related gene families, including the tonoplast aluminum-activated malate transporter, malate dehydrogenase, PEPC, and malic enzyme gene families (Yang ). Diversification of gene expression post-duplication suggests that natural selection is operating to create neo- or sub-functionalization.

Determination of gene function required for CAM using comparative genomics research

Although multiple CAM plant genomes have been sequenced, the bona fide function of the majority of these genes remains largely unknown and/or potentially misannotated. Comparative genomics effectively predicts genomic function and provides evolutionary insights across the phenomenal diversity of living systems (Butler ; Rogers, 2018). A rational strategy for elucidating gene function involved in CAM is to identify the changes in DNA/protein sequences and gene expression associated with the ontogenetic C3-to-CAM transition in constitutive CAM species, environmental induction/up-regulation of CAM expression in facultative CAM species, and differences between CAM and non-CAM species, as illustrated in Fig. 2. Here, several examples are presented to show how this strategy has been used to identify candidate genes related to CAM.

Fig. 2.

A comparative genomics framework for elucidating gene function associated with crassulacean acid metabolism (CAM). This framework integrates the ontogenetic C3-to-CAM transition in constitutive CAM species, environmental induction/up-regulation of CAM expression in facultative CAM species, and differences between CAM and non-CAM species.

In general, constitutive CAM is under ontogenetic control (Winter, 2019). In A. americana ‘Marginata’, the young leaf performs C3 photosynthesis and the mature leaf performs CAM photosynthesis (Abraham ). Recently, Yin performed comparative analysis of gene expression between mature and young leaf tissues of A. americana ‘Marginata’ and identified a co-expression cluster containing 1509 genes, which showed up-regulated expression in the mature leaves at night in comparison with mature leaves during the daytime and the young leaves during both night-time and daytime. This co-expression cluster contains several known CAM genes, including PPCK1, and it was therefore proposed that the genes in this module are related to the temporal reprogramming of metabolism underpinning CAM (Yin ). Also, Ping showed that the expression of a PPCK gene in Phalaenopsis aphrodite subsp. formosana was positively correlated with the progression from C3 to CAM at different development stages. Similarly, differential expression between non-photosynthetic (white base) and photosynthetic (green tip) tissue of the CAM plant A. comosus was revealed for genes encoding proteins (Ming ) as well as microRNAs (Wai ), suggesting that both protein-encoding genes and non-coding RNAs are involved in the developmental shift from C3 to CAM in this species.

It is agreed that facultative CAM is under environmental control (Winter, 2019). Recently, Brilhaus and Maleckova identified candidate transcription factors potentially involved in the regulation of the C3-to-CAM transition in the facultative CAM species Talinum triangulare, which is capable of reversible transitioning from C3 photosynthesis to weak CAM in response to drought stress. More recently, eight transcription factors were identified as potential candidates regulating the induction of CAM by water-deficit stress in another facultative CAM species, Mesembryanthemum crystallinum (Amin ). These comparative genomics analyses of stress-induced CAM versus C3 photosynthesis provide insights into the course of CAM evolution. However, additional effort will be needed to separate true CAM-related genes, and to establish their function, from the non-CAM genes that are concomitantly responsive to water-deficit stress. Comparative analysis of drought-responsive genes between facultative CAM species and closely related C3 species has the potential of narrowing down the list of candidate genes relevant to CAM in facultative CAM species. It is possible that the drought-induced genes in a C3 species closely related to a facultative CAM species have been co-opted to induce the CAM genes in the facultative CAM species. For example, existing regulatory proteins induced by drought in the C3 close relative may bind to promoters of the CAM genes in the facultative CAM species and bring about the induction of those genes just as they control other drought-responsive genes and pathways in the ancestral C3 species. Thus, some of the drought-induced genes in a C3 species may be the same genes that induce CAM in a closely related facultative CAM species. Furthermore, the candidate genes need to be tested using experimental approaches, such as loss-of-function analysis via knocking-out or knocking-down of the candidate genes in the facultative CAM species and gain-of-function analysis via overexpression of the candidate genes in the closely related C3 species, followed by evaluation of CAM-related traits (e.g. temporal pattern of CO2 uptake, malic acid accumulation) in the loss-of-function and gain-of-function mutant lines under drought stress. Stable genetic transformation is a bottleneck in experimental determination of gene function in plant species, for which robust transformation protocols have not been established or are difficult to establish. Therefore, it is critical to establish diverse CAM model species that are readily amenable to stable genetic transformation in order to be able to test the function of candidate CAM genes in a diverse range of CAM levels and phylogenetic origins (Hartwell ). For example, transcriptome-sequencing analysis of facultative the CAM species Clusia pratensis (Winter and Holtum, 2014) in comparison with its closely related C3 species C. multiflora, C. tocuchensis, or C. grandiflora (Zambrano ) under normal and drought conditions would help identify candidate genes responsible for the C3-to-CAM transition in C. pratensis. However, plant genetic transformation systems have not been reported for C. pratensis and its closely related C3 species yet, limiting our ability to experimentally validate computationally predicted CAM-related genes in C. pratensis. Therefore, there is an urgent need to establish efficient plant transformation systems for facultative CAM species and their closely related C3 species to facilitate the discovery of true CAM-related genes in facultative CAM species.

According to the relative contribution of dark CO2 fixation to the total 24 h net atmospheric CO2 fixation, constitutive CAM plants can be divided into two categories: strong constitutive CAM plants with dark CO2 fixation greatly exceeding CO2 fixation in the light under well-watered condition (K. Winter, personal communication) and weak constitutive CAM plants in which dark CO2 fixation accounts for less than ~5% of total carbon gain (Winter, 2019). In reviewing comparative genomics studies on the ontogenetic and environmental controls of CAM expression within individual CAM species, it is apparent that previous research has been focused on the genomic differences between weak and strong constitutive CAM species and between CAM and non-CAM species. Ontogenetic control of CAM expression (i.e. CAM appears only in the photosynthetic tissues passing a certain developmental stage) has been reported at the tissue level (e.g. irreversible transition from C3 to CAM during the early development of leaf tissue) or organismal level (e.g. progression from C3 to CAM during the development from young plantlets carrying mostly immature leaves toward adult plants) in both constitutive CAM species such as Kalanchoë spp. (Winter ; Winter, 2019) and facultative CAM species such as M. crystallinum (Herppich ). Also, CAM expression in both constitutive and facultative CAM species can be controlled or influenced by environmental factors, such as reduced soil water availability caused by drought or high salinity stress (Winter, 2019) and light quality (Ceusters ). For example, in the constitutive CAM species K. daigremontiana and K. pinnata, up-regulation of CAM expression can be stimulated by water stress and reversed by rewatering (Winter ). Among the light treatments (monochromatic wavelengths of blue, green, and red light) tested in the constitutive CAM plant Aechmea ‘Maya’, blue light was recently demonstrated to play a key role in regulating CAM expression in terms of stomatal responses, organic acid mobilization from the vacuole, and daytime decarboxylation (Ceusters ). This light regulation of CAM expression is potentially mediated by blue light receptors such as phototropin 2, which has been indicated to mediate blue light regulation of stomatal opening in Arabidopsis (Kinoshita ) and Kalanchoë (Yang ). Through comparative analysis of transcriptome profiles among closely related species performing strong CAM (Agave bracteosa and Manfreda spp.) and weak CAM (Beschorneria yuccoides and Polianthes tuberosa), Heyduk identified hundreds of co-expressed genes differing between weak and strong CAM, providing new insight into the gene function involved in the late stage (i.e. weak-to-strong CAM transition) of the CAM evolution path (Fig. 1).

Similarly, comparative analysis of gene expression between C3 and CAM species can provide insight into the gene function associated with the transition from C3 to CAM. So far, transcriptomic comparisons have focused on CAM species versus distantly related model C3 species. For example, the comparison of diel gene expression patterns between constitutive CAM species (e.g. Agave, Kalanchoë) and model C3 plant species (e.g. Arabidopsis, Solanum lycopersicum) has identified a number of candidate genes potentially involved in CAM metabolic processes, stomatal movement, and circadian rhythm (Abraham ; Moseley , 2019; Yin ). This type of comparative analysis potentially has a high false-positive rate because many of these candidate genes are potentially related to the general genetic divergence among distantly related CAM and C3 species. This limitation can be addressed by using the following approaches. First, comparative analysis of gene expression between closely related C3 and CAM species, such as CAM species in the genus Clusia (e.g. C. alata, C. hilariana, or C. rosea) versus closely related C3 species (e.g. C. multiflora, C. tocuchensis or C. grandiflora) (Zambrano ), CAM species in the genus Erycina (e.g. E. pusilla) versus the C3 species E. crista-galli (Heyduk ), and CAM species in the genus Yucca (Y. aloifolia) versus the C3 species Y. filamentosa (Heyduk ), can help elucidate gene function associated with the transition from C3 to CAM in a specific lineage while minimizing the false-positives caused by phylogenetic divergence. Recently, a comparative transcriptomic analysis of two closely related C3 and CAM species in the genus Erycina revealed candidate genes displaying significant differences in network connectivity, which were enriched with genes involved in light sensing and abscisic acid (ABA) signaling, including lov kelch protein 2 (involved in blue light sensing through regulation of light-induced protein degradation), phytochrome B/D (involved in red/far-red light sensing), time for coffee (responsible for the amplitude of the circadian clock but not thought to be directly involved in light signaling), protein phosphatase 2A (implicated in light signaling via dephosphorylation of phototropin 2), and seed longevity 1 (encoding RSL1, which acts as a master negative regulator of the ABA signaling pathway by targeting pyrabactin resistance 1 (PYR1) and PYR-like (PYL) ABA receptors for degradation) (Heyduk ). Second, comparative genomics analysis of multiple (i.e. more than one) distantly related CAM species, along with non-CAM species, can help uncover the gene function associated with phenotypic convergence among CAM plants. For example, it is widely accepted that there is one CAM-specific PEPC isoform (called PEPC1 herein), which is highly abundant and responsible for the CO2 fixation during the night-time (Abraham ; Boxall ). PEPC1 is sensitive to feedback inhibition by malate (Boxall ). Through genome-wide comparative analysis, Yang identified a new PEPC isoform (called PEPC2 herein) in two distantly related constitutive CAM plant species (K. fedtschenkoi and P. equestris), which share a single amino acid mutation (from a basic amino acid, R/K/H, to an acidic amino acid, D) in comparison with non-CAM plant species. This convergent single amino acid mutation appears to alleviate the feedback inhibition by malate. This PEPC2 gene was also shown to have a much higher transcript abundance during the dark period, suggesting the regulation of PEPC2 might be different from that of PEPC1 in K. fedtschenkoi and P. equestris (Yang ). Since the PEPC2 transcript was shown to have much lower abundance compared with the PEPC1 transcript in the whole leaf tissue of Kalanchoë (Yang ), it would be interesting to determine whether the PEPC2 transcripts are exclusively (or preferentially) expressed in other cell types of the leaf rather than the mesophyll cells. For example, PEPC is known to play important roles in the guard cells of plant leaves (Outlaw, 1990; Tarczynski and Outlaw, 1990; Outlaw ; Lawson, 2009; O’Leary ), so it is feasible that a second PEPC gene (PEPC2) detected in Kalanchoë leaves would be far less abundant than the CAM-specific PEPC gene (PEPC1). Future experiments, such as analysis of loss-of-function mutants and protein subcellular localization, will be needed to elucidate the function of the Kalanchoë PEPC2 gene.

New technologies for CAM comparative and evolutionary genomics research

With more genomics resources being created for CAM plants, it is becoming increasingly important to draw a deeper understanding of CAM evolution and gene function using computational and experimental approaches (Liu ). This deeper understanding can be facilitated through the use of new technologies developed over the past few years, including new computational tools and genome-editing technologies. Homology searches have been widely used in comparative genomics research. Recently, Price and Arkin (2017) developed an online tool called PaperBLAST that integrates sequence homology search and text mining to quickly find similar proteins with functional information buried in the scientific literature. Using a PaperBLAST search, more than 200 new candidate genes with potential roles in stomatal movement were identified in K. fedtschenkoi (Moseley ).

Protein sequence alignments have been widely used for CAM comparative genomics research, for phylogenetic analysis of gene families, and for the identification of protein sequence mutations in CAM plants in comparison with their orthologs in non-CAM species. Studying the structural changes caused by the amino acid mutations is helpful for understanding the impact of the protein sequence mutations on CAM evolution (Yang ). However, comparative analysis of three-dimensional protein structure is technically challenging compared with comparative analysis of one-dimensional protein sequences. To address this technical challenge, Babbitt recently developed a GUI-based pipeline, called DROIDS (Detecting Relative Outlier Impacts in Dynamic Simulations), for observing the biophysical effects of amino acid mutations in proteins and successfully applied this computational pipeline to the identification of disease-related functional changes in molecular dynamics caused by genetic mutation. This same computational tool has potential for investigating the role of protein sequence mutations in CAM evolution.

RNA interference (RNAi) has great potential for loss-of-function analysis in CAM plants (Dever ; Boxall ). For example, it was recently reported that RNAi-mediated down-regulation of two candidate genes involved in malate decarboxylation, encoding NAD-malic enzyme and pyruvate orthophosphate dikinase, respectively, in the constitutive CAM plant K. fedtschenkoi dramatically altered the temporal pattern of CAM CO2 fixation (Dever ), confirming that these two genes play important roles in CAM physiology in this species. More recently, Boxall demonstrated that RNAi-mediated reduction of PPCK1 transcript abundance in K. fedtschenkoi caused a dramatic decrease in dark phosphorylation of PEPC and consequently reduced dark period CO2 fixation, along with a loss of rhythmic expression pattern of multiple circadian clock genes, providing new insight into the function of PPCK1 in CAM plants. Precise genome engineering with CRISPR-Cas systems has shown great potential for functional genomics research in C3 and C4 plants, which has been summarized in several recent reviews (Liu ; Scheben and Edwards, 2018; Zhu ; Chen ). CRISPR/Cas-mediated gene knockout technology has several advantages over the RNAi-mediated gene knockdown approach: (i) it allows for complete shutdown of gene function; (ii) it is useful for functional characterization of intronic and intergenic regions; (iii) it has relatively low off-targeting activities; and (iv) it is feasible for characterization of non-coding RNAs functioning in the nucleus (Boettcher and McManus, 2015; Liu ; Rani ; Liu ). On the other hand, in comparison with CRISPR/Cas-mediated gene knockout technology, RNAi offers two important benefits: (i) it can generate multiple mutant lines with different degrees of knockdown, facilitating the quantitative analysis of gene expression in association with corresponding phenotypes (Liu ); and (ii) it is useful for studying the function of essential genes, for which the knockout mutations are lethal. Here, we focus on the utilization of CRISPR-Cas systems to understand CAM evolution and gene function.

(i) CRISPR/Cas-mediated knock-out can be applied to loss-of-function analysis of the candidate CAM-related genes predicted by CAM comparative and evolutionary genomics analysis (e.g. the genes involved in CO2 fixation, stomatal movement/development, leaf succulence, vacuole development, and the circadian clock). Recently, we established an efficient CRISPR/Cas9 genome editing system as an amenable approach for generating gene knockout mutants in the CAM model species K. fedtschenkoi (Liu ), demonstrating that CRISPR/Cas-mediated genome editing has the potential for comparative and evolutionary genomics research in CAM plants.

Functional characterization of the changes in DNA/protein sequences associated with CAM evolution can be achieved by the cytidine deaminase editor (Komor ) and adenine base editor (Gaudelli ) built on the CRISPR/Cas9 system. Recently, a CRISPR-guided DNA polymerase tool was engineered to continuously diversify all nucleotides within a tunable window length at user-defined loci in Escherichia coli (Halperin ), which potentially provides a new strategy for investigating the function of the candidate DNA/protein sequences shared in CAM species.

To study the impact of convergent shifts in diel patterns of gene expression in CAM species in comparison with C3 species, the CRISPR/Cas-mediated knock-in approach can be used to replace the promoters of CAM genes with the promoters of their orthologs in C3 plants. However, it is difficult to generate CRISPR/Cas-mediated knock-in mutants in plants. Several new strategies, such as DNA replicons (Baltes ), sequential transformation (Miki ), and RNA transcript-templated homologous recombination (Li ), could be tested in CAM plants to improve the efficiency of CRISPR/Cas-mediated knock-in.

Synthetic transcriptional activators and repressors, which are built on deactivated versions of Cas9 (dCas9) fused to transcriptional activation domains and repression domains, respectively, were recently used to control gene expression in non-CAM plants (Zalatan ; Papikian ). This dCas9-mediated gene regulation technology has great potential for testing the impact of gene expression changes on CAM evolution.

Comparing and contrasting the CAM gene function revealed by CRISPR-Cas mediated approaches and RNAi could provide complementary and comprehensive insights into CAM evolution.

Conclusion and perspectives

Over the past 5 years, comparative analyses of protein sequences and gene expression data have predicted hundreds of candidate genes that are potentially associated with metabolic processes, stomatal movement, and circadian rhythms in constitutive and facultative CAM plant species. Are these candidates, or just a subset of them, really involved in CAM-related biological processes? In other words, what is the minimum number of genes that are necessary and sufficient for converting from C3 to CAM photosynthesis? To determine whether the candidate genes are necessary for CAM photosynthesis, knockout mutants for each of these candidates need to be generated, followed by evaluating the impact of the mutants on CAM physiology. It would be technically challenging to generate knockout mutants for such a large number of candidate genes using traditional experimental approaches. In the foreseeable future, high-throughput CRISPR-based knockout systems need to be established to knock out the majority of the CAM candidate genes predicted in model CAM species (e.g. K. fedtschenkoi). To determine whether a set of candidate genes are sufficient for C3-to-CAM photosynthesis transition, various combinations of these genes should be overexpressed in model C3 species (e.g. Arabidopsis, tobacco) with appropriate promoters to ensure that the diel expression patterns of the transgenes recapitulate those in CAM plants. For example, the candidate genes involved in carboxylation, decarboxylation, stomatal movement, leaf succulence, vacuole development, and circadian rhythm in constitutive CAM plants as well as abiotic stress-responsive transcription factors in facultative CAM plants are targets for engineering CAM into C3 model species (Borland ; Ming ; Yang , 2017; Bräutigam ; Liu ; Moseley , 2019; Amin ).

Ecophysiological studies on various constitutive and facultative CAM plant species have generated several alternative hypotheses about the evolutionary path of C3-to-CAM progression. With more CAM-related genes being identified and characterized, the existing hypotheses for CAM evolution will be tested at the molecular level through comparative analysis of CAM-related genes in natural variants and mutant lines. With the rapid advance in genome and transcriptome sequencing of CAM plants, it is expected that comparative genomics research will provide new insights into the evolution of CAM, and the causal linkage between genes and CAM traits will be unraveled in the years to come. Application of new technologies and approaches may accelerate the progress in evaluating the feasibility of the two alternative models for CAM evolution: (i) the single-path model, in which the evolution of constitutive CAM was mediated by facultative CAM or (ii) the multi-path model, in which facultative CAM and constitutive CAM evolved independently from C3 photosynthesis (Fig. 1). These two models are not necessarily mutually exclusive. It is possible that both models are feasible depending on the phylogenetic position of a specific lineage and its evolutionary predispositions to follow a specific route to CAM during evolution.