Low internal air space in plants with crassulacean acid metabolism may be an anatomical spandrel

Abstract Crassulacean acid metabolism (CAM) is a photosynthetic adaptation found in at least 38 plant families. Typically, the anatomy of CAM plants is characterized by large photosynthetic cells and a low percentage of leaf volume consisting of internal air space (% IAS). It has been suggested that reduced mesophyll conductance (gm) arising from low % IAS benefits CAM plants by preventing the movement of CO2 out of cells and ultimately minimizing leakage of CO2 from leaves into the atmosphere during day-time decarboxylation. Here, we propose that low % IAS does not provide any adaptive benefit to CAM plants, because stomatal closure during phase III of CAM will result in internal concentrations of CO2 becoming saturated, meaning low gm will not have any meaningful impact on the flux of gases within leaves. We suggest that low % IAS is more likely an indirect consequence of maximizing the cellular volume within a leaf, to provide space for the overnight storage of malic acid during the CAM cycle.


INTRODUCTION
Saint Mark's Basilica in Venice is a location of important relevance to biologists.The layout of this building was used as a metaphor by Gould and Lewontin (1979) in their seminal paper critiquing adaptationist arguments in evolution.Specifically, Gould and Lewontin discussed the spandrels inside the central dome of the cathedral: the triangular spaces that occur when circular arches meet at right angles (Fig. 1).These spandrels, they argued, may at first glance appear to be deliberate spaces left to incorporate artwork within the church.However, Gould and Lewontin further note that spandrels are just a by-product of the architectural constraints imposed when building circular, intersecting domes.Put simply, you cannot build these sorts of domes without the existence of spandrels.Gould and Lewontin proposed that an analogous concept exists in evolution, arguing that not all traits observed in nature should be considered adaptations under selection.Some traits, behaviours or even ecological interactions are simply a by-product of other traits (Raia et al., 2010;Freeling, 2017;Valverde et al., 2018).For example, the human chin is often considered a spandrel, as it does not appear to play any functional role in survival (Pampush and Daegling, 2016).Instead, this protruding bone may be a by-product of muscles and teeth shrinking during the evolutionary history of Homo sapiens.More recently, discussion of spandrels has been extended to the field of functional trait ecology, which aims to explain ecological patterns and distributions using physiological, biochemical and morphological traits (Violle et al., 2007;Volaire et al., 2020;Griffith et al., 2022).Considerable debate is ongoing about which traits should be considered 'functional' -i.e. which traits truly aid the growth, reproduction and survival of an organism.Some authors have suggested that all traits are functional, and that any observed trait will either interact with the environment or another trait and affect survival, reproduction or vital rates (e.g.growth) (Sobral, 2021).However, this proposal overlooks the possibility that a trait may be a spandrel: rather than contributing to survival, it is simply a by-product of another characteristic which is itself under selection (Salguero-Gómez and Laughlin, 2021).

THE ARCHITECTURE OF PHOTOSYNTHETIC ORGANS
Most vascular plants achieve photosynthetic assimilation within leaves or photosynthetic stems.These organs contain photosynthetic parenchyma cells, whose main function is to capture light and use this energy to fix CO 2 into sugars.Between the mesophyll cells in leaves, internal air spaces allow the diffusion and circulation of gasses.There is considerable interspecific variation in the percentage of leaf volume consisting of internal air space (% IAS) (Slaton and Smith, 2002;Earles et al., 2018).In plants that do typical C 3 photosynthesis, low % IAS can limit photosynthesis as it minimizes mesophyll conductance (g m ), which controls the rate at which CO 2 can diffuse from the stomata to the chloroplasts (Baillie and Fleming, 2020;Cousins et al., 2020).CO 2 diffusivity is greater within the gas than the liquid phase of cells, and hence species with a greater % IAS can conduct CO 2 across their mesophyll more efficiently (Lawson et al., 2022).Additionally, the cell surface area exposed to IAS per leaf area (L mes /area) is also thought to limit g m in C 3 plants, by affecting the rate at which CO 2 dissolves into the liquid phase, across the cell wall and into the chloroplast.Ultimately, reductions to g m decrease photosynthetic assimilation rates in C 3 leaves, as CO 2 is slower to reach the chloroplasts (von Caemmerer, 2000).
Whilst the relationship between % IAS and photosynthetic assimilation in C 3 plants is well accepted, this association is less straightforward in species that do other forms of photosynthesis (Knauer et al., 2022).Approximately 7 % of all vascular plants do a form of photosynthesis called crassulacean acid metabolism (CAM), which is distinct from the more common C 3 pathway (Winter, 2019;Winter et al., 2021).Unlike C 3 photosynthesis, in which CO 2 assimilation occurs only in the day, CAM plants can exhibit four distinct phases of gas exchange over a 24-h period (Fig. 2).The majority of CO 2 assimilation in CAM plants occurs at night (phase I), via the enzyme phosphoenolpyruvate carboxylase (PEPC).PEPC fixes CO 2 into oxaloacetic acid, which is then converted to malic acid, transported into the vacuole and stored overnight (Borland et al., 2014).The following day, this malic acid is decarboxylated, thereby regenerating CO 2 which can enter the Calvin-Benson-Bassham cycle (Ceusters et al., 2021).In the morning (phase II) stomata are open, allowing RuBisCO to increasingly fix CO 2 entering the leaf from the atmosphere as well as CO 2 generated from the decarboxylation of malic acid.However, during the hotter, drier middle of the day (phase III), CAM plants shut their stomata, and the Calvin-Benson-Bassham cycle relies entirely on malic acid decarboxylation for CO 2 input.During phase III, some species exhibit a net efflux of CO 2 , due to imperfectly closed stomata (Friemert et al., 1986).Finally, in the evening (phase IV), when malic acid reserves are depleted, stomata re-open and RuBisCO fixes atmospheric CO 2 .The four phases of gas exchange in CAM plants are depicted in Fig. 2.There are two primary benefits conferred by CAM (Holtum, 2023).First, the diurnal decarboxylation of malic acid during phase III allows photosynthesis to occur without reliance on atmospheric CO 2 , meaning CAM plants can keep their stomata shut during the hotter, low-humidity portions of the day (Borland et al., 2015).Consequently, CAM plants conserve water, which allows them to survive in drier ecological niches (Bone et al., 2015;Leverett et al., 2021;Schweiger et al., 2021).The second benefit of CAM is that intercellular and chloroplastic CO 2 concentrations (c i and c c , respectively) are elevated during phase III, compared to ambient atmospheric conditions (Owen and Griffiths, 2013;Heyduk et al., 2019;Boxall et al., 2020).Hence, CAM is often described as a carbon-concentrating mechanism (CCM).By elevating c c during the decarboxylation phase, CAM plants both increase the rate of RuBisCO-mediated carboxylation and decrease photorespiratory rates by reducing the wasteful oxygenation reaction also catalysed by RuBisCO (Shameer et al., 2018).CAM is often found in fleshy, succulent leaves and/or stems (Borland et al., 2018;Males, 2018).Large succulent cells contain greater vacuole space, which aids the CAM cycle as more malic acid can be stored overnight (Töpfer et al., 2020;Leverett et al., 2023a).As well as large cells, another anatomical trait found in CAM plants is low % IAS (Smith and Heuer, 1981;Borland et al., 2018;Males, 2018).It has been proposed that unlike C 3 photosynthesis, low % IAS actually aids CAM plants.
Here, we critically assess the possible ways in which low % IAS could benefit CAM plants and put forward an alternative hypothesis: that this anatomical trait is best considered a spandrel, occurring as a consequence of tightly packing cells into a leaf to maximize the storage space for malic acid.

Argument 1: Do benefits arise from low g m during phase III of CAM?
The hypothesis that low % IAS is beneficial to CAM plants is based on the assumption that this anatomical configuration will lower g m , thereby slowing the rate that CO 2 diffuses out of cells and ultimately leaks out of the leaf, during phase III.Under this scenario, low % IAS is thought to help leaves elevate c c and enhance the CCM function of CAM.This hypothesis was first proposed by Nelson et al. (2005), and has been largely accepted by much of the CAM research community (Nelson and Sage, 2008;Ceusters and Borland, 2010;Ripley et al., 2013;Barrera-Zambrano et al., 2014;Heyduk et al., 2016aHeyduk et al., , b, 2020;;Borland et al., 2018;Males, 2018;Edwards, 2019;Leverett, 2019;Niechayev et al., 2019;Gilman and Edwards, 2020;Fradera-Soler et al., 2021).Nelson et al. (2005) observed that mesophyll cell size did not correlate with % IAS across 18 phylogenetically diverse CAM species, leading them to conclude that the latter was unlikely to be an indirect consequence of developing larger cells to facilitate malic acid storage.However, a lack of any correlation between % IAS and mesophyll cell size does not necessarily mean that the low % IAS is independent of malic acid storage capacity.It is worth noting that there is no a priori reason that CAM plants need large mesophyll cells.If CAM requires adequate storage space for malic acid, this could equally be achieved through the development of a greater number of small cells.Larger cells may have simply been preferred during the evolution of CAM because it requires less carbon investment in cell walls (John et al., 2017).If one were to consider storage space for malic acid through the lens of total cellular volume per leaf area (X vmax ), then X vmax can be approximated by: where T is the thickness of mesophyll tissue, excluding tissue that does not do CAM, such as hydrenchyma (Smith et al., 1987;Leverett et al., 2023a).Hence, there are two ways of increasing the malic acid storage capacity for CAM: by increasing mesophyll thickness (often achieved with larger cells) and by reducing % IAS.The importance of % IAS for X vmax can be appreciated when considering the variation across the Bromeliaceae.Earles et al. (2018) found that in this family, T was ~0.5 mm and % IAS was ~10 % in CAM species, yielding an X vmax of 0.45.However, if CAM species had % IAS values comparable to those seen in C 3 species (~40 %), X vmax becomes 0.3.Hence a lower % IAS contributes to a 50 % higher X vmax in CAM species of Bromeliaceae.This illustrates that the lack of any correlation between % IAS and mesophyll cell size does not necessarily mean that low % IAS is independent of malic acid storage capacity.Consequently, it is possible that low % IAS in CAM plants is not functionally important, but is instead a result of maximizing cell volume within leaves, to enhance malic acid storage capacity of leaves (Maxwell et al., 1997;Heyduk et al., 2016aHeyduk et al., , 2020;;Earles et al., 2018).Given that % IAS and X vmax are often closely linked, it is difficult to compare the effect each trait has on the physiology of CAM plants.Most comparative physiology approaches would not be sufficient to determine if low % IAS aids the CAM cycle by reducing g m , or if it is a spandrel that occurs from maximizing the volume of cells.However, several models have been made that describe the physiology of CAM plants (Owen and Griffiths, 2013;Bartlett et al., 2014;Hartzell et al., 2018Hartzell et al., , 2020;;Töpfer et al., 2020;Wang et al., 2023).Models can be used to artificially alter physiological and anatomical parameters, in ways that may not occur in nature.This was done by Owen and Griffiths (2013), in a model that captures all four phases of CAM.Owen and Griffiths simulated changes to g m and X vmax , independently of one another, something that would rarely occur in natural systems.They showed that increasing X vmax resulted in greater CO 2 assimilation during phase I of CAM, demonstrating that greater malate storage capacity results in a direct benefit to photosynthetic assimilation rates in CAM plants.In contrast, reducing g m had no discernible effect on CO 2 efflux during phase III.The implication of this finding was that low % IAS (and the reduced g m it confers) does not substantially affect c i during phase III.Consequently, we suggest that the more likely hypothesis is that low % IAS is an indirect consequence of maximizing the cellular volume within leaves, in order to increase X vmax and elevate nocturnal CO 2 assimilation rates.Despite the Owen and Griffiths model having been published a decade ago, its implications for the role that low % IAS and g m plays in CAM photosynthesis have been largely ignored.We suggest that the most likely explanation for why low g m will not substantially impact CO 2 efflux during phase III is that stomatal resistance is far greater than mesophyll resistance during this time (note that resistance is the inverse of conductance).Low % IAS is expected to directly reduce conductance through the air space (g IAS ) whilst also indirectly decreasing L mes /area, which will reduce the conductance across the cell wall (g w , see Fig. 3) (Nelson and Sage, 2008;Knauer et al., 2022).Both g IAS and g w will contribute to g m as a whole.Thus, it is possible to predict the effect that low % IAS will have on the overall diffusion of CO 2 , by considering g m .For CO 2 to exit a leaf during phase III of CAM, it must first travel from the cytoplasm though the mesophyll air space, then out of stomata.This can be conceptualized as two resistors in series, the first with conductance, g m and the second with conductance g sc (stomatal conductance to CO 2 ).Therefore, the total conductance to CO 2 between the sites of carboxylation and the leaf surface (g msc ) can be derived as (Jones, 2013): According to eqn (2), changes to g m will have little to no impact on phase III CO 2 losses (governed by g msc ) if g m remains substantially higher than g sc .Put simply, when g m ≫ g sc , changes to the former will have little effect on net efflux of CO 2 during phase III.We surveyed experimentally measured values of g m , g sc .max (i.e. when stomata are open) and g sc .min (leakiness through the cuticle and closed stomata) from the literature on CAM plants (Table 1).In Kalanchoë daigremontiana (Crassulaceae), g sc .min was three orders of magnitude smaller than g m , meaning even extremely large changes to the latter would have almost no impact on CO 2 efflux during phase III.
In Aechmea fendleri (Bromeliaceae), we could not find an estimate of g sc .min from the literature.However, the conductance of internal air space to CO 2 (g IAS ) in this species is higher than g sc .max , which suggests that stomatal conductance is sufficiently small to prevent low % IAS from limiting CO 2 efflux during phase III.Finally, in the C 3 -CAM intermediate species Clusia minor, g sc .max was higher than g m .However, estimates of g sc .min were four to six orders of magnitude lower than g m , meaning stomatal closure prohibits g m from limiting CO 2 efflux during phase III.Together, these data indicate that stomatal limitations to CO 2 efflux obviate any limitations g m can impose on the efflux of CO 2 during phase III of CAM.The thick waxy cuticles and low stomatal densities found in most CAM species are likely to result in g sc .min values being orders of magnitude lower than g m (Kerstiens, 1996;North et al., 2019;Kubásek et al., 2023).Indeed, a recent survey in the genus Aeonium found that conductance across closed stomata was significantly lower in CAM species than for C 3 relatives (T.Messerschmid, personal comm.).Consequently, low % IAS is very unlikely to have any meaningful impact on c i or the carbon balance of phase III of CAM.Taken together, these data suggest that low % IAS in CAM plants is best considered a spandrel.Future work should prioritize estimating g m , g IAS and g sc .min in more constitutive CAM plants, grown under the same conditions, to understand the relative contribution each trait has on g msc .
Argument 2: Do benefits arise from low g m during phase II and IV?
In addition to considerations of phase III, it has been suggested that low g m may aid CAM by helping to trap CO 2 during phase II, when CO 2 entering the Calvin-Benson-Bassham cycle is derived from both the decarboxylation of malic acid and the influx through open stomata (Cousins et al., 2020).This hypothesis is similar to that considered in Argument 1: that low g m would reduce the efflux of CO 2 out of cells and aid CAM as a CCM.During phase II, the measured net leaf CO 2 exchange, A n , will be determined by: where V c and V o are the carboxylation and oxygenation rates of RuBisCO, R d is the respiratory rate and D c is the rate at which CO 2 is generated from the decarboxylation of malate.Thus, during phase II, a net positive A n indicates that the rate of RuBisCO carboxylation is higher than the combined rate of CO 2 generation from photorespiration, respiration and malate decarboxylation.Furthermore, because stomata are open during phase II, A n can also be given by the product of the stomatal-mesophyll conductance to CO 2 , g msc , and the concentration gradient of CO 2 from the atmosphere, c a , to the site of carboxylation, c c , i.e.: where g msc is formulated according to eqn (2).Malate decarboxylation will elevate c c values, although to what degree is unknown.However, because A n is positive and g msc is finite, c c must Densely packed, succulent mesophyll Low %IAS Cuticle Epidermis Chloroplasts Fig. 3.The pathway of CO 2 diffusion away from the chloroplasts can be represented as a series of resistors with conductances: g c (from the cell wall to the site of carboxylation), g w (across the cell wall), g IAS (through the intercellular air space) and g sc (across the stomata).The mesophyll conductance, g m , can be conceptualized as comprising g c , g w and g IAS in series such that The CO 2 concentrations at the site of carboxylation, the intercellular air space and the atmosphere are given by c c , c i and c a , respectively.Cell vacuoles are not shown.
be less than c a .This leads to an increase in CO 2 uptake through the stomata, with the CO 2 concentration in the intercellular space, c i , remaining below the atmospheric CO 2 concentration, c a .Therefore, during phase II, a lower mesophyll conductance will decrease g msc and should decrease net carbon assimilation.This makes sense when considering that diffusion is bidirectional.If A n is positive, then more CO 2 must be diffusing from the atmosphere to the chloroplasts than from the site of malate decarboxylation to the atmosphere.Changes to % IAS would have the same impact to the diffusive resistance for both fluxes.
Because more CO 2 is moving into the leaf than out, lowering g m will result in a net reduction to A n .Consequently, whilst lower g m will help trap CO 2 originating from the decarboxylation of malate, this will not incur any benefit during phase II because leaf gas exchange measurements indicate inward CO 2 diffusion is greater than leakage of decarboxylated CO 2 .Hence, the benefit of CO 2 trapping will be outweighed by the assimilatory penalty incurred to the CO 2 entering the leaf through open stomata.This argument also applies to other periods where A n is positive, such as phases I and IV of CAM: whenever A n is positive, reductions to g m will cause CO 2 assimilation to decrease.It is worth noting that this has additional consequences for Argument 1: even if low g m does reduce CO 2 efflux during phase III, this would only have an adaptive benefit if the CO 2 saved were greater than the cost it would incur from reduced A n during phases I, II and IV.Therefore, future investigations should compare closely related constitutive CAM species or populations to ask the following questions: (1) Does low % IAS result in less CO 2 loss during phase III of CAM, and is this truly independent of X vmax ?(2) Are the reductions to phase III CO 2 efflux greater than the losses incurred to A n during phases I, II and IV?

IMPLICATIONS FOR CAM EVOLUTION AND CAM BIODESIGN
Low % IAS has previously been considered an important anatomical adaptation for CAM photosynthesis.However, we propose that this trait may be a spandrel: a by-product of maximizing the volume of cells within a given volume of leaf.
Treating low % IAS as a spandrel will be important to the discussion on the evolution of CAM.This metabolic adaptation is thought to have evolved from C 3 origins in at least 38 plant families, making it one of the most remarkable cases of convergent evolution in nature (Winter et al., 2021;Gilman et al., 2023).
However, the intermediate steps on the evolutionary trajectory from C 3 to CAM remain enigmatic.It has been suggested that an antagonistic relationship exists between the anatomical configuration best suited for C 3 photosynthesis (higher % IAS, higher g m ) and CAM (lower % IAS, lower g m ).Such an antagonism would make it difficult to evolve a strong CAM phenotype from C 3 origins, as it would limit the viable phenotypic space that C 3 -CAM intermediate species could inhabit (Edwards, 2019).However, if low % IAS does not confer any appreciable benefit to CAM, this antagonism would not exist.Therefore, % IAS may not constrain evolutionary trajectories across the C 3 -CAM continuum.
From a biotechnological perspective, treating low % IAS as a spandrel may also be important.Substantial efforts are underway to bioengineer the CAM pathway into C 3 crops, to benefit from the drought tolerance this metabolic pathway confers (Borland et al., 2015).To achieve this goal, the anatomy of leaves in a host plant must be optimized to ensure bioengineered CAM pathways function efficiently.To this end, succulent, transgenic lines of Arabidopsis thaliana have been developed, which will ensure sufficient space is available for nocturnal storage of malic acid (Lim et al., 2018(Lim et al., , 2020)).Whilst these transgenic lines also have lower % IAS than wild-type plants, the correlation between % IAS and cell size is weak, meaning one trait can be prioritized when selecting the ideal host line for CAM biodesign (Lim et al., 2020).We suggest that cell size, and not % IAS, should be the criterion with which hosts are chosen, as the latter is unlikely to result in any benefits to the net carbon balance of synthetic CAM pathways.This will ensure the optimal anatomy is selected to aid with bioengineered CAM.Finally, whilst current work is attempting to express a constitutive CAM cycle in C 3 species, future projects will probably try to bioengineer facultative CAM physiology into crops (Borland et al., 2018).Facultative CAM plants can switch their photosynthetic physiology from C 3 to CAM in response to water limitations.Facultative CAM plants are therefore able to achieve higher CO 2 uptake rates associated with C 3 photosynthesis when growing in optimal conditions, whilst also benefiting from the water-conserving properties of CAM when water is scarce (Winter and Holtum, 2014).However, evidence from Clusia suggests that the anatomy of facultative CAM species often more closely resembles that of C 3 species than constitutive CAM species (Barrera-Zambrano et al., 2014;Leverett et al., 2023b).The rates of C 3 photosynthesis in well-watered facultative CAM plants are likely to be limited by g m , meaning higher % IAS would benefit facultative CAM plants that are under optimal conditions.Furthermore, having a high % IAS would confer no downside when CAM is facultatively induced, as this trait will not limit the rate of CO 2 efflux during phase III, when the stomata are shut.In addition, higher g m may increase the rate of photosynthetic assimilation during phases I, II and IV, as CO 2 can more efficiently reach reaction centres.Consequently, we propose that the optimal anatomical configuration for biodesigned facultative CAM is one where % IAS is high, and g m is comparable to values seen in C 3 species.This would benefit plant productivity when water is available.In addition, when drought has resulted in the facultative engagement of CAM, high % IAS could result in elevated CO 2 assimilation during phases I, II and IV, without any considerable changes to CO 2 efflux during phase III.

CONCLUSIONS
The anatomy of CAM plants is often associated with low % IAS, which has led some to suggest that this trait is under selection.We propose an alternative hypothesis: low % IAS is a spandrel resulting from CAM plants maximizing the volume of cells within a leaf or cladode.

Fig. 1 .
Fig. 1.Triangular spandrels (grey arrows) are present in architectural designs when circular arches meet at right angles.

Fig. 2 .
Fig. 2. The four phases of gas exchange common to many CAM plants.Phase I: CO 2 assimilation occurs at night, via PEPC, and carbon is stored as malic acid.Phase II: stomata open in the morning and CO 2 is increasingly fixed by RuBisCO.Phase III: stomata shut during the middle of the day and malic acid decarboxylation produces CO 2 to enter the Calvin-Benson-Bassham cycle.Note that during phase III some CO 2 can leak out of imperfectly closed stomata.Phase IV: malic acid reserves are depleted, and stomata reopen; CO 2 is fixed by RuBisCO.
a Maxwell et al. e *