Stomatal Biology of CAM Plants1[CC-BY]

Recent advances in the stomatal biology of CAM plants are reviewed, and key opportunities for future progress are identified.

physiological syndrome that has evolved independently in numerous land plant lineages. CAM plants are of great ecological significance, and there is increasing interest for their water-use efficiency and drought resistance. Integral to the improvement in water-use efficiency that CAM affords is a unique pattern of stomatal conductance, distinguished by primarily nocturnal opening and often extensive diurnal flexibility in response to environmental factors. Here, we assess how recent research has shed new light on the functional biology of CAM plant stomata and integration within the broader physiology and ecology of succulent organisms. Divergences in stomatal sensitivity to environmental and endogenous factors relative to C 3 species have been a key aspect of the evolution of functional CAM. Stomatal traits of CAM plants are closely coordinated with other leaf functional traits, and structural specialization of CAM stomatal complexes may be of undiagnosed functional relevance. We also highlight how salient results from ongoing work on C 3 plant stomatal biology could apply to CAM species. Key questions remaining relate to the interdependence between stomatal and mesophyll responses and are particularly relevant for bioengineering of CAM traits or bioenergy crops to exploit enhanced water-use efficiency and productivity on marginal land. With the increasing availability of powerful analytical tools and the emergence of new model systems for the study of the molecular basis of physiological traits in CAM plants, many exciting avenues for future research are open to intrepid investigators. CAM is a celebrated example of a convergent physiological syndrome (i.e. a characteristic combination of traits), having evolved independently on numerous occasions across the land plants . Furthermore, thanks in part to their ability to withstand multiple, synergistic stressors (Lüttge, 2010), CAM plants have successfully invaded diverse environmental spaces ranging from deserts to cloud forests. In many tropical and subtropical vegetation types, CAM is a dominant ecophysiological syndrome, and CAM plants represent at least 6% of higher plant species richness (Dodd et al., 2002).
The physiological mechanisms and ecological significance of the gas exchange rhythms of plants performing CAM have been the subject of curiosity and investigation for not just decades, but centuries (De Saussure, 1804;Heyne, 1815;Osmond, 1978;Ting, 1987;Faak, 2000). The quintessential feature of CAM is nocturnal primary carbon assimilation by the enzyme phospho-enol-pyruvate carboxylase (PEPC), producing malic acid that is stored in mesophyll cell vacuoles and subsequently decarboxylated during the light period to provide CO 2 for refixation by Rubisco . While a few lineages are capable of performing CAM in tissues lacking stomata, including some aquatic plants with leaves with no stomata ("astomatous"; Keeley, 1998) and epiphytic orchids with astomatous chlorophyllous roots (Goh et al., 1983), in most cases, CAM involves the delivery of CO 2 to the mesophyll via stomata that are open in the dark . Nonnegligible nocturnal stomatal conductance is increasingly recognized as an important physiological phenomenon in many C 3 plants (Zeppel et al., 2012;de Dios et al., 2013;Forster, 2014;Matimati et al., 2014;Zeppel et al., 2014;Cirelli et al., 2016;Resco de Dios et al., 2016), but stomata of CAM plants displaying primarily nocturnal CO 2 assimilation clearly must differ from those of C 3 plants in their responsiveness to environmental and endogenous stimuli.
The global CAM flora combines great ecological diversity with a wide variety of evolutionary backgrounds, and comparative studies of variation in the stomatal biology of different CAM lineages allow two overarching questions to be distilled. First, what characteristics unite the functional biology of CAM plant stomata? Were there multiple evolutionary routes to the same phenomenology, or do all CAM plants share the same molecular and metabolic basis for stomatal behavior? Second, how does variation in stomatal form and function among CAM species underpin physiological adaptation to the wide range of environmental niches these plants have come to occupy?
Researchers have adopted a multiplicity of approaches to shed light on these questions, spurred both by the enduring appeal of CAM as a "curiosity" (Osmond, 1978) and the rapidly growing interest in the application and engineering of CAM plants for bioenergy production (Borland et al., 2011(Borland et al., , 2014(Borland et al., , 2015Owen and Griffiths, 2014;DePaoli et al., 2014;Yang et al., 2015). Simultaneously, the wider field of stomatal biology has experienced a renaissance in recent years, with numerous advances being made through both empirical and theoretical work. Although this research has generally been carried out in a C 3 context, lessons can be carried through to the CAM world. Here, we provide a general synthesis of current understanding of CAM stomatal biology and identify key opportunities for future research.

PATTERNS OF STOMATAL CONDUCTANCE
The four classical phases of CAM, driven by changes in carbon metabolism, coincide with changes in stomatal conductance across the diurnal cycle ( Fig. 1; Osmond, 1978). Stomatal conductance is typically highest during the dark period (Phase I), in association with nocturnal CO 2 assimilation by PEPC. During the dark period, mesophyll factors are often more important in limiting the rate of nocturnal assimilation than is stomatal conductance (Winter, 1985;Winter et al., 1985). Around dawn, there is often a spike in stomatal conductance and some direct fixation of CO 2 by Rubisco (Phase II), which continues to fix CO 2 released by decarboxylation of malic acid behind closed stomata during most of the light period (Phase III). During the late afternoon, if environmental conditions are favorable there may be a period of stomatal opening with direct Rubisco-mediated fixation of CO 2 (Phase IV). However, this canonical pattern of gas exchange is subject to a large amount of interspecific, intraspecific, and intraindividual variation. One of the most remarkable features of CAM is its plasticity in response to environmental variability. The expression of the classical phases of CAM is modulated in response to recent and current environmental conditions (Dodd et al., 2002;Owen and Griffiths, 2013). Under low water availability and high evaporative demand, for instance, Phase IV stomatal opening may be completely abolished.
Additionally, two frequently observed modes of CAM do not conform to the textbook four-phase gas exchange profile: "CAM cycling" and "CAM idling" (Sipes and Ting, 1985). CAM cycling involves the nocturnal operation of respiratory recycling and diurnal stomatal opening for direct Rubisco-mediated assimilation and most often occurs as a facultative trait in C 3 -CAM or "weak CAM" species (Silvera et al., 2010). Meanwhile, under CAM idling, stomata remain closed throughout the day and night, with a proportion of respiratory CO 2 being refixed. CAM idling is often induced under extreme seasonal drought stress in "strong CAM" species, maximizing water retention (Silvera et al., 2010). This capacity for close environmental tracking on both diurnal and seasonal bases maximizes integrated water-use efficiency and is therefore an important contributor to the ecological success of CAM plants in stressful habitats (Fig. 1).

STOMATAL SENSITIVITY TO ENDOGENOUS STIMULI: CONTROL OF, AND BY, CAM
Stomata and mesophyll cell processes may be controlled by distinct circadian clocks in C 3 species (Hubbard and Webb, 2016), and understanding the interplay between these cycles could provide insights for the coordination of CAM, as well as the interplay between responses to internal and external signals via metabolite feedback, internal CO 2 availability (C i ), and environmental cues (Fig. 1).
The circadian rhythm of CAM plants involves the same system of clock genes as have been intensively studied in Arabidopsis (Arabidopsis thaliana; Boxall et al., 2005;Hubbard and Webb, 2016) and controls diurnal oscillations in physiological processes including photosynthetic enzyme activity (Nimmo, 2000;Hartwell, 2005). However, there is evidence for an important role for metabolite control of the temporal dynamics of the CAM cycle. For instance, manipulation of key decarboxylation and metabolite regeneration processes in Kalanchoë fedtschenkoi had a direct disruptive effect on the mesophyll circadian clock (Dever et al., 2015). Furthermore, reducing the capacity of CAM leaves to synthesize malic acid at night (by removal of external CO 2 supply) showed that associated reductions in metabolite concentrations could override circadian control of PEPC kinase (Borland et al., 1999). However, the extent that the guard cell circadian cycle is synchronized with, or driven by, the mesophyll CAM cycle, remains to be determined.
Additional insights have been gained from combined measurements of gas exchange and carboxylation enzyme coregulation under continuous light. In Mesembryanthemum crystallinum, the timing of circadian rhythms of stomatal conductance, CO 2 assimilation, and Rubisco/PEPC continued to be synchronized across light and dark cycles (Davies and Griffiths, 2012). However, stomatal conductance was lower when Rubisco carboxylation predominated at the end of the Plant Physiol. Vol. 174, 2017 551 light period and higher when PEPC carboxylation predominated at the end of the dark periods, perhaps suggesting that guard cells are responding to the extent of CO 2 drawdown and intercellular CO 2 concentration (C i ; Davies and Griffiths, 2012). Sensing CO 2 concentrations has long been implicated in both C 3 and CAM stomatal movements; intuitively, responding to C i would seem to be the signal most likely to regulate the inverse stomatal cycle associated with CAM. At the beginning of Phase I of CAM, stomatal opening is thought to be driven by reduced C i when PEPC activity increases at dusk (Wyka et al., 2005;Griffiths et al., 2007;von Caemmerer and Griffiths, 2009). In the morning, stomatal closure is then reinforced by the decarboxylation of stored malate during Phase III. This, coupled with respiration, can cause C i to increase up to 100 times atmospheric concentration (Cockburn, 1979;Spalding et al., 1979). The reopening of stomata to initiate Phase IV is associated with the end of malic acid breakdown and, hence, internal CO 2 limitation.
The responsiveness of CAM stomata to changing ambient CO 2 transients was investigated in relation to degree of leaf succulence and commitment to the CAM cycle (von Caemmerer and Griffiths, 2009). The stomata of the more succulent Kalanchoë daigremontiana were more responsive to a CO 2 transient reduction at night, whereas stomata in the less succulent Kalanchoë pinnata were more responsive during daytime Phase IV gas exchange. When CO 2 uptake and malic accumulation Thicknesses of arrows indicate relative importance of factors in typical species of either photosynthetic pathway. The response of guard cells (GC) to blue light appears to be weaker or absent in many CAM species (e.g. Tallman et al., 1997), while strong sensitivity to leaf-air VPD occurs in many CAM plants (e.g. Lange and Medina, 1979). Leaf (or whole-plant) water status, often acting on stomatal aperture in an ABA-dependent manner, may have a stronger effect in CAM species (Jewer et al., 1981). Intercellular CO 2 concentration (C i ) is a key driver of stomatal rhythms in CAM plants (e.g. Wyka et al., 2005), and this is intimately linked to the importance of metabolite control in CAM (Dever et al., 2015). While the circadian oscillator still plays a role in shaping CAM stomatal rhythms (von Caemmerer and Griffiths, 2009), it is perhaps of less direct importance than in C 3 plants. Structural modifications are common in stomata of both C 3 and CAM species, but specialized stomatal complex morphologies with large subsidiary cells appear to be overrepresented among CAM plants (see "Stomatal Structure-Function Relationships").
were reduced overnight, and subsequent C i regeneration lowered during Phase III, stomata still closed and showed little instantaneous response to CO 2 transients, suggesting that circadian control of stomata remains a key factor controlling the CAM cycle in both species (von Caemmerer and Griffiths, 2009). However, there is still a lack of clarity in defining the interplay between circadian inputs from guard cells and mesophyll metabolism and how sensing of C i and metabolites is transduced by stomata in CAM plants. The major advances in our understanding of the mechanism of CO 2 sensing and regulation of stomatal conductance in C 3 plants (Chater et al., 2015;Engineer et al., 2016) provide an excellent springboard for exploration of the role of equivalent genetic systems in CAM species. Abraham et al. (2016) have already demonstrated that there is a concerted shift in the temporal expression of components of CO 2 signaling pathways in the constitutive CAM species Agave americana relative to C 3 Arabidopsis. The generality of this observation among other CAM systems should now be explored and the regulatory mechanisms further elucidated.

STOMATAL RESPONSES TO EXTERNAL STIMULI
In addition to circadian control of stomatal and mesophyll processes, environmental tracking by CAM plant stomata is mediated by the integration of endogenous and exogenous signals by guard cells, as in C 3 species (Assmann and Jegla, 2016).
The role of blue light in the stomatal movements of CAM plants has also not been fully resolved (Kinoshita, 2017). While there is some evidence for the involvement of blue light signaling in the regulation of stomatal conductance and malate decarboxylation in CAM bromeliads (Ceusters et al., 2014) and for the induction of CAM in Clusia minor (Grams and Thiel, 2002), other studies performed with facultative CAM plants have concluded that blue light regulates stomatal conductance of these plants only when they are in the C 3 mode (Lee and Assmann, 1992;Tallman et al., 1997). Moreover, the results of transcriptomic analysis of the constitutive CAM plant A. americana were not consistent with a role for stomatal regulation by blue light (Abraham et al., 2016). This apparent divergence in stomatal regulation in different CAM lineages could hint at the existence of multiple mechanistic routes to CAM-like stomatal function.
Both leaf water potential and the humidity of the leaf microenvironment also affect stomatal conductance. Declining leaf water potential is a powerful driver of stomatal closure in C 3 plants ( Rodriguez-Dominguez et al., 2016). Although comparative data are quite limited, succulent CAM plant stomata tend to close at much higher (less negative) water potentials than those of co-occurring C 3 plants (Osmond, 1978), consistent with evidence that succulent plants tend to avoid, or be isolated from, drought stress (Nobel, 1988; J. Males and H. Griffiths, unpublished data). Complete stomatal closure can therefore occur throughout both the light and dark periods (CAM idling).
Malate has been proposed as a mesophyll to guard cell signal in the regulation of stomatal aperture in response to mesophyll turgor and light-dark transitions in C 3 plants (Araújo et al., 2011;Costa et al., 2015), while oxaloacetate has been shown to be an effective inhibitor of guard cell anion channel activity (Wang and Blatt, 2011). The involvement of organic anions in stomatal regulation has interesting implications for CAM plants, in which malate can accumulate to high concentrations during Phase I (Osmond, 1978). The importance of abscisic acid (ABA), which is synthesized and mobilized in roots and shoots in response to declining water potential, in regulating stomatal closure in CAM plants, compared to C 3 plants, remains to be determined (Cutler, 2017;Jezek and Blatt, 2017). Jewer et al. (1981) suggested that stomata of CAM plants might be hypersensitive to ABA, which would be consistent with strategies for avoiding soil water deficits (tissue water potentials usually . 21 MPa), water storage, and rapid recharge in succulent tissues. Recent progress in our understanding of the role of ABA in the evolution of stomatal responses should be brought to bear on CAM plants (Negin and Moshelion, 2016). The debate over the origins of signaling pathways for both ABA and CO 2 continues, and contrasting observations in ferns (which do contain CAM lineages; Ong et al., 1986;Winter et al., 1986) await resolution (McAdam and Brodribb, 2012;compare Chater et al., 2015;Franks and Britton-Harper, 2016).
An apparent feed-forward response of transpiration to rising leaf-air vapor pressure deficit (VPD), in which stomata seem to respond directly to humidity rather than indirectly via leaf water status, has been observed in some CAM lineages, with important consequences for assimilation rates and water-use efficiency under contrasting humidity regimes (Lange and Medina, 1979;Osmond et al., 1979;Martin and Siedow, 1981;von Willert et al., 1985;Lüttge et al., 1986;Herppich, 1997). Epiphytic CAM species might be expected to show particularly high levels of stomatal sensitivity to VPD, given the special adaptive value this would have in highly water-limited epiphytic environments (see discussion of integrated leaf traits below). Indeed, in C 3 plants, stomatal sensitivities to VPD and leaf water potential are often strongly correlated with leaf or petiole hydraulic conductances and their sensitivity to tissue water potential (Brodribb and Jordan, 2008;Ocheltree et al., 2013Ocheltree et al., , 2014Klein, 2014;Tombesi et al., 2014;Bartlett et al., 2016). The mechanisms underlying stomatal sensitivity to VPD remain a controversial and active area of research, with the possibility of liquidand/or vapor-phase signals being involved alongside ABA synthesis and signaling within guard cells (Peak and Mott, 2011;Bauer et al., 2013;Buckley and Mott, 2013;Mott and Peak, 2013;McAdam et al., 2016). Because of the potentially significant metabolic and signaling interactions between guard cells and the mesophyll, integrated investigation of stomatal sensitivity and the dynamic responses of the critical extravascular component of leaf hydraulic conductance in CAM (and C 3 ) species is highly desirable Trifiló et al., 2016). Analysis of the spatiotemporally dynamic expression patterns of aquaporins and of possible interactions between stomatal physiology and mesophyll osmotic properties could be especially fruitful (Pou et al., 2013;Martorell et al., 2015).
A final factor that has been demonstrated to influence stomatal conductance in CAM plants is temperature, with optimal CAM activity usually associated with narrow and relatively low (usually ;15-25°C) nocturnal temperature windows (Yamori et al., 2014). Both thermoperiodic effects (Ting et al., 1967) and instantaneous leaf temperature effects (Nobel and Hartsock, 1979) have been reported. Given the known importance of nocturnal leaf temperature for the efficiency of malate synthesis and decarboxylation (e.g. Neales, 1973;Moradshahi et al., 1977;Nobel and Hartsock, 1984), water-use efficiency should be maximized through the regulation of stomatal conductance in line with temperature.

GUARD CELL METABOLISM
Guard cell metabolism in C 3 , C 4 , and CAM plants continues to be a fast-paced area of research with many critical questions awaiting resolution (Daloso et al., 2016;Santelia and Lunn, 2017). The similarities between guard cell metabolism in C 3 plants and the metabolism of mesophyll cells of CAM plants are striking, which led Cockburn (1981) to suggest that a transfer of guard cell-like metabolism to mesophyll cells was a central event in evolutionary origins of CAM. More recent work has highlighted the importance of organic acids in C 3 guard cell function (e.g. Wang and Blatt, 2011;Penfield et al., 2012;Daloso et al., 2015;Medeiros et al., 2016).
Controlled ion fluxes are fundamental to the operation of stomatal movements (Chen et al., 2012;Minguet-Parramona et al., 2016;Eisenach and de Angeli, 2017;Jezek and Blatt, 2017). In comparing the day-night transcriptomic profiles of C 3 Arabidopsis and the constitutive CAM plant A. americana, Abraham et al. (2016) showed that there was a coordinated shift in the temporal expression patterns of key ion channels in A. americana. Notably, orthologous vacuolar chloride channel genes displayed reciprocal expression in the C 3 and CAM species, which could help to drive appropriate charge balancing.
The presence of Rubisco in the guard cells of some CAM plants needs further investigation in the context of the emerging role of guard cell photosynthesis in the regulation of stomatal conductance in C 3 plants (Madhavan and Smith, 1982;Azoulay-Shemer et al., 2015). Tallman (2004) suggested that guard cell photosynthesis could be supplied with large amounts of CO 2 from the mesophyll during Phase III of CAM, establishing a strong sink for NADPH and thus inhibiting the degradation of guard cell endogenous ABA, which promotes stomata closure (Lind et al., 2015). In this way, guard cell photosynthesis in CAM plants could assist in the maintenance of negligible diurnal stomatal conductance during the light period. Santelia and Lawson (2016), citing earlier work carried out by Pantoja and Smith (2002), recently highlighted the absence of the correlation between malate currents across the guard cell tonoplast and cytosolic calcium concentrations across CAM species that would be expected if they shared a uniform regulatory mechanism. This apparent diversity in stomatal physiology could have important consequences for our understanding of the evolution of complex syndromes like CAM. Further empirical studies of this topic are needed to advance our understanding of the imposition of daytime stomatal closure in CAM plants. Cell-specific perturbation of metabolic function offers an exciting opportunity in this respect .

COORDINATION OF STOMATAL TRAITS WITH LEAF TRAIT NETWORKS
CAM species have rarely been included in analyses of leaf economic trait variation, partly because succulence is one trait which uncouples leaf mass-based relationships (Grubb et al., 2015;J. Males and H. Griffiths, unpublished data). However, in a survey of leaf economic, anatomical, and hydraulic traits in the Bromeliaceae, we found that CAM bromeliads tended to show lower stomatal density and conductance as well as lower leaf hydraulic conductance, photosynthetic capacity, and nutrient content, and higher leaf mass per unit area (J. Males and H. Griffiths, unpublished data). Variation in stomatal traits appears to be accommodated within a network of coordinated leaf traits in CAM species in the same way as has been observed in C 3 plants (Reich et al., 1997(Reich et al., , 1999Wright et al., 2004Wright et al., , 2005Donovan et al., 2011;Vasseur et al., 2012;Díaz et al., 2016). Recent modeling and empirical studies have highlighted the importance of the alignment of variation in stomatal, xylem, and veinal traits in angiosperms for optimal physiological function (Brodribb et al., 2013(Brodribb et al., , 2016Fiorin et al., 2016;Carins Murphy et al., 2016;Scoffoni et al., 2016). It would be particularly interesting to explore the degree of coordination between Phase I (nighttime) and Phase IV (daytime) stomatal and mesophyll conductances in CAM plants. Although few data are available, it is expected that mesophyll conductance is generally low in CAM plants due to their succulent anatomy with tight cellpacking (Maxwell et al., 1997;Nelson and Sage, 2008). Campany et al. (2016) recently showed that coupled responses of stomatal and mesophyll conductances to light improved carbon gain during sunfleck events in shade leaves of a Eucalyptus species. Similar effects are likely to be important in CAM epiphytes of the humid tropics with sunfleck-driven carbon economies.

STOMATAL STRUCTURE-FUNCTION RELATIONSHIPS
CAM has arisen in a wide range of taxonomic and morpho-anatomical backgrounds, and this is reflected in the various stomatal complex morphologies found in different CAM lineages. When variation is considered among the angiosperms at the family level, using the APG IV classification (Angiosperm Phylogeny Group, 2016) and anatomical data from the DELTA database (Watson and Dallwitz, 1992), the proportional occurrence of different stomatal complex morphologies shows several potentially important differences between CAM and C 3 lineages. None of the monocot families with CAM elements display anomocytic stomata (lacking subsidiary cells), whereas 26% of exclusively C 3 monocot families do. Tetracytic stomata (four subsidiary cells) are nearly twice as common in CAM families as in C 3 families. Among the dicots, anomocytic stomata are also less common in CAM families, and there are relatively more CAM families with paracytic stomata (two subsidiary cells). The overrepresentation of CAM in families with more specialized stomatal complexes in both monocots and dicots has not been investigated from a functional perspective. However, it is well established that the presence of subsidiary cells in C 3 and C 4 species can enhance the kinetics of stomatal movements (Franks and Farquhar, 2007), and systematic differences in stomatal kinetics and sensitivity may occur between CAM species with contrasting stomatal morphologies. Empirical and theoretical work in the C 3 context also suggests that stomatal size could be an important determinant of the rapidity of stomatal movements (Drake et al., 2013;Lawson and Blatt, 2014;Raven, 2014), although this relationship may be modulated by guard cell morphology (McAusland et al., 2016). These trait linkages are potentially of great evolutionary and ecological importance and could easily be tested for in CAM plants. It is interesting to note that among the few fern lineages to have evolved CAM, modified polocytic and pericytic stomatal complexes occur, wherein the guard cell pair is surrounded either completely or partially by one or two subsidiary cells (e.g. Patel et al., 1975;Sen and Hennipman, 1981).
When compared with their nearest C 3 relatives, CAM lineages show no consistent differences in guard cell ultrastructure (Faraday et al., 1982) but do tend to display a shift toward lower stomatal densities and lower maximal conductances in CAM plants (Ting et al., 1972;Kluge and Ting, 1978;Gibson, 1982;Zambrano et al., 2014;J. Males and H. Griffiths, unpublished data). These reductions have widely been interpreted as adaptive xeromorphic traits in their own right, but there is now accumulating evidence for a developmental constraint that generates a robust negative relationship between stomatal density and the sizes of guard cells and mesophyll cells (Brodribb et al., 2013). Since CAM is dependent on the presence of highly vacuolate succulent cells for malate storage, low stomatal densities could be a necessary trade-off. Further investigation of the coordination of stomatal traits, cell sizes, succulence, and perhaps genome sizes (Beaulieu et al., 2008) could prove illuminating.

STOMATA ON THE RUGGED CAM ADAPTIVE LANDSCAPE
CAM is now often discussed as a continuum of intergrading and flexible photosynthetic modes rather than a monolithic, discrete trait (Silvera et al., 2010;Winter et al., 2015). The existence of a wide range of CAM types and the occurrence of evolutionary reversions from CAM to C 3 (Teeri, 1982a(Teeri, , 1982bSilvera et al., 2009;Givnish et al., 2014) is a reflection of a rugged adaptive landscape with multiple peaks. While the description of the C 3 -C 4 adaptive landscape as "Mount Fuji-like" (Heckmann et al., 2013) is a simplified abstraction, there are convincing accounts of the demonstrable increases in fitness associated with each step between full-C 3 and full-C 4 metabolism in independent C 4 origins (Christin et al., 2011(Christin et al., , 2013Griffiths et al., 2013;Heckmann et al., 2013;Sage et al., 2013;Schlüter and Weber, 2016). In the absence of the wealth of phylogenetic and physiological information enjoyed by the C 4 community, and despite the possibility that C 4 and CAM represent alternative evolutionary pathways from similar starting points (Edwards and Ogburn, 2012), the picture for CAM is far murkier (Hancock and Edwards, 2014). Succulence has been identified as an anatomical prerequisite for CAM (Sage, 2002;Zambrano et al., 2014;Heyduk et al., 2016), but beyond this there is little clarity regarding the relative timing of the acquisition of component traits of the CAM syndrome or the extent to which different types of CAM could represent independent adaptive peaks. In particular, the involvement of stomatal innovation in convergent origins of CAM is unclear. How does the capacity for stomatal flexibility vary among the C 3 sister taxa of CAM lineages? During evolutionary transitions from C 3 to CAM, do any less obvious changes in stomatal biology occur prior to the appearance of the inverse stomatal rhythm? Is the answer to this question the same for lineages that have only evolved weak CAM (CAM cycling) as for those that have evolved strong CAM? Concerted efforts to improve phylogenetic resolution in critical lineages in which C 3 -to-CAM transitions have occurred, a more accurate diagnosis of "cryptic" low-level CAM, and targeted surveys of stomatal physiological traits and molecular biology in representative taxa would all be important preliminary steps toward unraveling these long-standing evolutionary puzzles.

CONCLUSIONS AND FUTURE PERSPECTIVES
CAM is a major ecophysiological syndrome that has been repeatedly identified as providing high potential for sustainable production under climate change (Borland et al., 2011(Borland et al., , 2014(Borland et al., , 2015Owen and Griffiths, 2014;Yang et al., 2015). Harnessing this potential is contingent upon a comprehensive understanding of the underlying physiology of CAM. Recent work has contributed to our knowledge of how stomatal specialization is involved in the unique metabolic flexibility and water-use efficiency afforded by CAM, while insights gained from work on the stomatal biology of non-CAM plants can also be reinterpreted from a CAM perspective. However, there is still much to be learned about the functioning of CAM stomata (see Outstanding Questions). One promising route for future research will be to make use of known C 3 -CAM intermediates and facultative CAM species as tools for exploring the molecular changes associated with the commencement of CAM stomatal rhythms (Winter and Holtum, 2014;Brilhaus et al., 2016). The identification of gradients in the relative contributions of C 3 and CAM along the linear leaves of C 3 -CAM intermediate monocot species is another naturally occurring system ripe for further investigation (Popp et al., 2003;Freschi et al., 2010). Increasingly sensitive technologies will improve the ease of in situ and ex situ physiological characterization (e.g. Barkla and Rhodes, 2017), and robust transcriptomic methodologies will be crucial for elucidating the molecular genetic basis of divergences in stomatal function along the CAM continuum and under variable environments. Finally, the integration of recently developed physiological models of CAM (Owen and Griffiths, 2013;Bartlett et al., 2014;Hartzell et al., 2015) with more detailed models of stomatal conductance could be a powerful way of exploring the significance of variation in stomatal traits for carbon gain and wateruse efficiency.