Why is C4 photosynthesis so rare in trees?

We review possible physiological, anatomical, and evolutionary limitations that may explain the rarity of C4 photosynthesis in trees and discuss how the C4 trees in Euphorbia are exceptions to this.


Introduction
C 4 photosynthesis arises from anatomical and biochemical modifications to the ancestral C 3 photosynthetic machinery that serve to concentrate CO 2 around Rubisco (Box 1). This CO 2 -concentrating mechanism (CCM) acts to elevate CO 2 assimilation, while functionally increasing the apparent specificity of Rubisco for CO 2 , over O 2 , thus minimizing oxygenation and resultant photorespiration (Tcherkez et al., 2006). It follows that the evolution of C 4 photosynthesis is favoured by environmental conditions that would promote high rates of photorespiration in C 3 species, namely low ambient CO 2 concentrations, warmth, bright light, aridity, and salinity (Chollet and Ogren, 1975;Ehleringer et al., 1991;Sage et al., 2018). Since C 4 photosynthesis was first observed >50 years ago (Kortschak et al., 1965;Hatch and Slack, 1966), numerous studies have attempted to elucidate exactly which modifications are typically required to assemble the components of C 4 physiology (Box 2). To travel the path of C 4 evolution, an ancestral C 3 progenitor arrived in an environment selective for a C 4 benefit (Ehleringer et al., 1997;Sage, 2001a;Sage et al., 2018), and, starting from an initial set of genetic and anatomical pre-adaptations (Monson, 2003;Christin et al., 2013Christin et al., , 2015Griffiths et al., 2013), evolved developmental and genetic modifications (Stata et al., 2016;Moreno-Villena et al., 2018;Dunning et al., 2019a;Lundgren et al., 2019), navigated energetic constraints (Bellasio and Lundgren, 2016), and underwent progressive optimization (Rondeau et al., 2005;Christin et al., 2009)-or, in some cases, potentially 'cheated' this lengthy final step via horizontal gene transfer (Christin et al., 2012a, b;Dunning et al., 2019b). Given that some version of this path has been repeatedly travelled nearly 70 times by diverse plant lineages spanning a wide range of lifeforms and ecological niches (Sage et al., 2011a), it seems unusual that only a single group of true trees (i.e. defined here as tall, perennial, woody lifeforms with secondary growth) has evolved C 4 leaves (Pearcy and Troughton, 1975; Table 1). These exceptional trees are members of Euphorbia, a global genus of Euphorbiaceae spanning the semideserts of East Africa to the rainforests of the Pacific Islands, and encompassing growth forms from herbs to xerophytic stem succulents to trees of up to 30 m in height (Horn et al., 2012).
Box 1. C 4 photosynthesis arises from both anatomical and biochemical modifications to the ancestral C 3 photosynthetic system There are several anatomical and biochemical differences that arise during the transition from the C 3 photosynthetic system to the C 4 CO 2 -concentrating mechanism (CCM). In C 3 plants, the majority of chloroplasts (and associated Rubisco) are localized to the mesophyll, which is largely exposed to ambient CO 2 and O 2 concentrations. The featureless nature of CO 2 and O 2 makes them enzymatically hard to distinguish, such that Rubisco has catalytic affinity for both molecules, and will catalyse the carboxylation and oxygenation of RuBP. Because the oxygen availability to Rubisco is high in C 3 plants, oxygenation and subsequently photorespiration occur at high rates, especially in high-temperature and low-CO 2 conditions. Some C 3 plants have evolved higher specificity of Rubisco for CO 2 over O 2 ; however, this comes at the cost of slower catalytic turnover (Tcherkez et al., 2006).
The C 4 CCM allows plants to avoid this specificity-efficiency trade-off by increasing CO 2 concentrations at the Rubisco active site, leading to an apparent increase in the specificity of Rubisco for CO 2 . In C 4 plants, CO 2 is biochemically shuttled from the mesophyll into the bundle sheath via the carbonic anhydrasephosphoenolpyruvate carboxylase (PEPC) system, which builds up CO 2 concentrations around Rubisco within bundle sheath cells, drastically reducing the incidence of oxygenation and increasing net carbon assimilation.
To facilitate the C 4 cycle, C 4 plants have higher bundle sheath to mesophyll area ratios (Hattersley, 1984), often via increased density of vascular bundles . In addition, the connectivity of the mesophyll and bundle sheath cells is enhanced in C 4 plants by an increased density of plasmodesmata at the cell interface, which allows for the increased flux of metabolites that is required for a functional C 4 cycle (Danila et al., 2016).

Mesophyll
Bundle Sheath Euphorbia is the only known genus to contain plants using crassulacean acid metabolism (CAM), C 3 , C 3 -C 4 , and C 4 photosynthetic types (Yang and Berry, 2011). The C 4 lineage in Euphorbia, the largest single C 4 lineage among the eudicots, is found within subgenus Chamaesyce, a subclade of Euphorbia that includes C 3 and C 4 species, as well as C 3 -C 4 evolutionary intermediates (Sage et al., 1999a(Sage et al., , 2011bYang and Berry, 2011;Box 3). Subgenus Chamaesyce underwent a radiation on the Hawaiian Islands, resulting in 27 taxa that all use C 4 photosynthesis, as indicated by δ 13 C values spanning -14.5‰ to -12.0 ‰ (Pearcy and Troughton, 1975;Sporck, 2011). The Hawaiian species in Chamaesyce (hereafter referred to as Hawaiian Euphorbia, still recognizing that a separate colonization event led to the origin of a C 3 Hawaiian tree not in Chamaesyce, Euphorbia haeleeleana) include a variety of growth forms, from sub-shrub to tree, which exist across a diverse Box 2. Key developments in understanding C 4 evolution There are numerous evolutionary steps on the path from C 3 to C 4 (Stata et al., 2019) . To understand this complex evolutionary progression, it is most useful to examine it in study systems that contain individuals across the entire spectrum of photosynthetic types from C 3 to C 4 . While there are many fully C 3 and C 4 Euphorbia species, there are only two known C 3 -C 4 intermediates, and this limits the use of Euphorbia as a model for understanding C 4 evolution. However, study of the eudicot genus Flaveria has been crucial in understanding the evolutionary transitions from C 3 -C 4 to C 4 in particular (Monson and Moore, 1989), while studies of the grass Alloteropsis semialata provide unparalleled insight into C 4 evolution owing to its extreme intraspecific photosynthetic diversity, which reduces the confounding effects of long divergence times in phylogenetic analyses and species comparisons Dunning et al., 2019a). Key developments in these systems have started to pick apart the anatomical and biochemical components required to construct a functional C 4 photosynthetic system. The remarkable convergence of the C 4 trait across the plant kingdom means that the findings from Flaveria and A. semialata can be applied to other distantly related C 4 lineages, such as Euphorbia, despite differences in life history, to understand which stages on the path to C 4 might conflict with the tree habit.
• High expression of key genes is important in the establishment of a weak C 4 cycle. Moreno-Villena et al.
(2018) demonstrated that highly expressed genes in grasses, and also possibly in Flaveria, were preferentially co-opted for C 4 photosynthesis regardless of tissue specificity. The importance of high expression levels of C 4 cycle genes was further shown by the observed increases in expression of the key C 4 genes phosphenolpyruvate carboxykinase (PEPCK) and phosphenolpyruvate carboxylase (PEPC) across the C 3 to C 3 -C 4 transition (Dunning et al., 2019a). These genes underwent duplication and a resultant dosage-dependent increase concurrently with their co-option for C 4 photosynthesis in Alloteropsis semialata (Bianconi et al., 2018). • C 4 anatomy can evolve via a single developmental change: an increase in vein density. Lundgren et al. (2019) demonstrated that an increase in vein density driven by proliferation of minor veins is sufficient, given necessary anatomical pre-conditions, to produce a functional C 4 leaf anatomy and create an evolutionary entry to more complex C 4 syndromes in A. semialata. Christin et al. (2013) present the necessary anatomical pre-conditions (or 'enablers') in grasses as a high (>15%) proportion of bundle sheath tissue (combination of a short distance between bundle sheaths and large bundle sheath cells), which, when combined with environmental changes, facilitated the emergence of the C 4 trait. • A reduction in chloroplast numbers and increased chloroplast size is associated with changes in C 4 metabolic activity. Stata et al. (2016) showed that across C 3 , C 3 -C 4 , and C 4 species in the genus Flaveria, chloroplast number and coverage of the mesophyll cell periphery increase with increased strength of C 4 metabolism, while increased C 4 cycle strength was associated with increased chloroplast size. The reduced chloroplast volume at the mesophyll cell periphery and associated increased cytosolic exposure to the atmosphere could enhance diffusion of CO 2 to PEPC, thus facilitating the incorporation of CO 2 into the C 4 metabolic cycle. • Lateral gene transfer is an important force in C 4 evolution, spreading functional genes among grasses.
A total of 59 genes in the genome of A. semialata were laterally acquired from other grasses, including those known to be involved in C 4 photosynthesis (Dunning et al., 2019b;Olofsson et al., 2019).
While these anatomical and biochemical modifications that assemble the C 4 trait were selected for by increased fitness in hot, dry environments, they have been retained by the entire Hawaiian Euphorbia across multiple transitions from open habitats into the moist forest understorey following the colonization of the Hawaiian Islands, and multiple transitions to the tree habit (Yang et al., 2018). Considering these recent developments in our understanding of C 4 evolution, alongside the ecology and biogeography of the Hawaiian radiation, is key to understanding the overall evolutionary progression of the lineage. range of environments, from bright and arid habitats to mesic and wet forest understoreys (Table 1). Though there are some relatively shade-tolerant C 4 monocots (e.g. Lundgren et al., 2015), shade-tolerant C 4 eudicots are rare (Sage et al., 1999b) and thus the ubiquity of C 4 photosynthesis across the diverse habitats of Hawaiian Euphorbia is surprising (Pearcy and Troughton, 1975;Sage, 2016). Given that the entire Hawaiian radiation uses C 4 photosynthesis, it is likely that the progenitor for this radiation was also a C 4 species that arrived on the Hawaiian Islands. This idea is supported by recent phylogenetic work that suggests that the closest relatives of Hawaiian Euphorbia are in fact C 4 herbs from the southern USA, Mexico, and/or the Caribbean, and that the woody state evolved after arrival on the Hawaiian Islands (Yang and Berry, 2011;Yang et al., 2018;Box 3).
There has been periodic interest in Hawaiian Euphorbia, and the wider rarity of C 4 trees (e.g. Pearcy and Troughton, 1975;Ehleringer, 1978;Pearcy, 1983;Sporck, 2011;Sage, 2014;Sage and Sultmanis, 2016;Yang et al., 2018); however, there has been little definitive progress towards understanding why C 4 trees are indeed so rare. Given that C 4 trees do exist, there cannot be any fundamental incompatibility between C 4 photosynthesis and the Herbs to sub-shrubs tree habit, or any physiological explanation for the reduced competitive ability of a C 4 tree versus a C 3 tree that is true under all conditions. Therefore, in explaining the rarity of C 4 trees, all the factors that could limit, but not preclude, the evolution of the two syndromes in the same species must be considered, whether that evolution is via adaptation of a C 4 progenitor to an environment inhabited by trees, as is likely to be the case for Hawaiian Euphorbia, or via an existing tree traversing the adaptive landscape from C 3 to C 3 -C 4 to C 4 (Yang and Berry, 2011). This review will consider the key steps on the path to C 4 photosynthesis, where these steps might conflict with the tree lifeform, and argue that Hawaiian Euphorbia present a unique opportunity to study the evolution of the C 4 trait in trees as a target for future research.

C 4 trees may perform poorly under a closed canopy
Rates of photorespiration increase in warm environments, making the C 4 pathway-which largely avoids photorespiration-superior to C 3 photosynthesis in plants with similar lifeforms (Ehleringer, 1978). As such, C 4 grasses frequently dominate in areas with warm climates where trees cannot grow, for example due to high levels of disturbance, while C 3 forestsand thus canopies-establish in warm areas where conditions are such that trees thrive. The theory that follows is that C 4 trees have failed to become widespread forest species due to their poor performance under canopies, where conditions are cool, shady, and often enriched in CO 2 (Sage et al., 1999a). However, this theory may not hold true given insights into the physiological performance of Euphorbia species in the Hawaiian forest understorey (Robichaux and Pearcy, 1980;Pearcy et al., 1985) and, more recently, C 4 photosynthesis under low light canopy conditions (Bellasio and Griffiths, 2014;Sage, 2014).

Reduced quantum yield as a limitation
Ehleringer (1978) proposed an early hypothesis that the quantum yield of photosynthesis, defined as the rate of photosynthesis relative to that of photon absorption, is important in determining the distribution of C 4 species, especially grasses. Maximum quantum yield is inherently lower in C 4 plants than in C 3 plants due to the greater energy requirements of the C 4 system, though the quantum yield of C 3 plants declines with increasing temperature while that of C 4 plants remains constant (Long, 1999;Monson, 1999). If below-canopy temperatures are sufficiently cool that the additional energy requirement of the C 4 system is greater than the light energy lost to photorespiration in an energetically inexpensive C 3 plant, then the quantum yield of the C 3 plant would be greater than that of a theoretical C 4 tree, and that tree would probably be outcompeted. Indeed, for Atriplex species native to grassland, desert, and coastal strand habitats, the temperature at which the quantum yields of C 3 and C 4 species are equal is 30 °C, at least at atmospheric CO 2 (at that time 325 ppm) and O 2 concentrations (Ehleringer and Björkman, 1977). Quantum yield also varies with CO 2 concentration, life history, and C 4 biochemical subtype. The lower quantum yield of eudicots compared with monocots may contribute to the relative scarcity of shadetolerant C 4 eudicot herbs compared with forest-shade grasses (Monson, 1999). Similarly, the higher quantum yields of plants using the NADP-ME biochemical subtype of C 4 photosynthesis (such as Hawaiian Euphorbia) compared with those using the NAD-ME subtype (Ehleringer and Pearcy, 1983;Pearcy and Ehleringer, 1984) may partially explain the shade tolerance of the understorey C 4 tree Euphorbia herbstii (formerly E. forbesii), whose quantum yield equals that of an equivalent C 3 tree at a leaf temperature of 22-23 °C (approximately the same value as the mean midday leaf temperature at the site where these plants were collected) (Robichaux and Pearcy, 1980). Direct comparison between the quantum yields of C 3 and C 4 species, however, does not adequately address the question of whether or not quantum yield could limit the evolution of a C 4 tree: the quantum yields of intermediate species on the path from C 3 to C 4 must be considered. In the incipient C 3 -C 4 phases of C 4 evolution in Flaveria, the poor integration of the C 3 and C 4 cycles causes futile cycling in the C 4 assimilation of CO 2 and thus reduced quantum yields (Monson et al., 1986;Stata et al., 2019). Inefficient transfer of CO 2 from the C 4 to the C 3 cycle may create an 'adaptive trough', expressed through reduced quantum yields in C 3 -C 4 species, compared with fully coupled C 3 or C 4 taxa, which could act as a barrier to the evolution of C 3 -C 4 traits in species native to shady habitats (Monson, 1989). Thus, the limitations of the C 3 -C 4 intermediate state could make it difficult for the C 4 pathway to evolve in a tree under a forest canopy. However, this limitation does not apply where the transition to the forest understorey occurred subsequently to the evolution of the full C 4 trait, as was likely to have been the case for Hawaiian Euphorbia (Yang and Berry, 2011). Sage et al. (1999a) proposed that C 4 plants may be maladapted to shady understorey environments due to their inefficient utilization of sunflecks, which represent the primary source of light available under the canopy. However, sunfleck use does not seem to be a limiting factor in Euphorbia, as the C 4 tree E. herbstii is as efficient in utilizing sunflecks as a comparative C 3 tree (Pearcy et al., 1985). Similarly, study of the C 4 grass maize shows that, while it responds more slowly to short sunflecks, it otherwise has a similar sunfleck use efficiency to C 3 crop species (Krall and Pearcy, 1993), suggesting that C 4 plants may not be inherently limited by poor sunfleck exploitation.

Inability to meet the increased energy demands of C 4 in the shade
Under low light, C 4 plants suffer increased 'bundle sheath leakiness'-the rate of diffusion of CO 2 out of the bundle sheath relative to that of phosphoenolpyruyvate (PEP) carboxylation (Kromdijk et al., 2008). This is driven by a slow down in the C 4 carboxylation process due to a reduction in ATP availability under low light. Thus, in the low-light forest interior, the higher leakiness would limit carbon gain and lead to poorer performance of a theoretical C 4 tree. However, maize plants grown under diffuse light can acclimate and thereby reduce leakiness, possibly by allocating proportionally more energy to C 3 cycle activity to reduce CO 2 overcycling, optimizing scarce ATP resources, and then trapping a greater proportion of CO 2 in the bundle sheath (Bellasio and Griffiths, 2014).
Disjunct veins, which shade-tolerant Hawaiian Euphorbia species possess (Herbst, 1971), may be another mechanism adapted to tolerate shade. This modification to leaf anatomy allows these species to have a low density of functional veins (i.e. those that are connected to the vascular network) to reduce leaf costs, as is typical of shade species (Sporck, 2011), while establishing islands of bundle sheath tissue to increase the relative bundle sheath tissue area and maintain the close proximity of mesophyll and bundle sheath cells required for a functional C 4 system (Box 1). Therefore, it seems that optimization of C 4 physiology in combination with modifications to leaf anatomy should allow C 4 plants to thrive under the canopy, including C 4 trees such as E. herbstii and E. rockii (Table 1). However, not all C 4 lineages may be equally primed to adopt these modifications.

Evolutionary factors shaped the pathway to C 4 trees
There are several key historical factors and evolutionary 'opportunities' that have played a role in shaping the evolution of the C 4 trait, and subsequently the tree habit, in Hawaiian Euphorbia.

Evolution of the C 4 trait in Euphorbia
First, it is important to consider the age of the eudicot C 4 lineages and the timing of C 4 evolution relative to historical climatic changes. C 4 eudicots are not overall younger than C 4 monocots, and all lineages of C 4 plants evolved in the low CO 2 atmosphere that has shaped plant evolution over the last 30 million years (Christin et al., 2011). This low CO 2 atmosphere was probably a key evolutionary opportunity for Euphorbia and is associated with the evolution of at least 17 independent CCMs, which are mostly CAM but also include the C 4 lineage (subsect. Hypericifoliae) of Euphorbia subg. Chamaesyce (Horn et al., 2014;Sage, 2016).
Secondly, relatively short generation times in Euphorbia, owing to rapid flowering and high levels of reproductive output per plant, favour the comparatively fast evolution of traits, probably including those associated with C 4 photosynthesis, as well as the rapid accumulation of duplicated and neofunctionalized genes as a resource for C 4 evolution (Monson, 2003;Emms et al., 2016;Bianconi et al., 2018;Box 2). However, this rapid generation time probably offers less potential evolutionary benefit than lateral gene transfer, which is increasingly recognized as an important force in shaping C 4 evolution in the monocots, but has not been documented in C 4 eudicots (Christin et al., 2012a, b;Dunning et al., 2019b;Olofsson et al., 2019;Box 2).
Thirdly, lifeform may have also played a role in the evolutionary potential of these plants. The ancestor of the C 4 lineage in Euphorbia was likely to have been herbaceous, while all 16-21 independent origins of CAM in Euphorbia occurred in woody ancestors (Horn et al., 2014). While this is only a single example, it suggests that the woody ancestor of Euphorbia may have needed to undergo a transition to the herbaceous lifeform prior to the evolution to C 4 photosynthesis, and perhaps the evolution of C 4 is less favourable than that of CAM in a woody species. Furthermore, the herbaceous lifeform of the ancestor of Hawaiian Euphorbia facilitated its dispersal to the Hawaiian Islands (Yang and Berry, 2011). Indeed, much of the Hawaiian woody flora evolved from herbaceous ancestors, which had greater dispersal ability to reach the remote Hawaiian Islands (Carlquist, 1970;Panero et al., 1999;Eggens et al., 2007;Dunbar-Co et al., 2008). Before European contact and the influx of invasive species, Hawaii had an environment with available niches along with topographic heterogeneity providing barriers to gene flow (Givnish et al., 2009). Thus, C 4 trees or their progenitors (none of which are able to dominate the forest canopy) did not need to invade established C 3 communities or areas of high disturbance that typically occlude the tree lifeform, so providing an evolutionary opportunity for a C 4 ancestor to transition to the tree habit.
Finally, Euphorbia have a remarkably high degree of variation in morphological characteristics, level of adaptive plasticity, and species richness compared with other plant lineages of similar age (Horn et al., 2012). In particular, there are several characteristics of the ancestral Euphorbia that may be synergistically associated with the origin of the C 4 lineage, and thus may have facilitated C 4 evolution. The combination of plagiotropic branches and a distichous leaf arrangement maximize the leaf area exposed to sunlight. The co-evolution of C 4 photosynthesis with these growth traits would maximize photosynthetic rates and minimize photorespiration rates in high-light, hightemperature environments. In addition, the high adaptive plasticity of Euphorbia may have facilitated the evolution of further adaptations in progenitors of the C 4 trees, for example the development of shade-tolerant leaves, circumventing constraints that the C 4 state places on phenotypic plasticity (Herbst, 1971;Robichaux and Pearcy, 1980;Sage and McKown, 2006;Horn et al., 2012;Lundgren et al., 2014). It is also worth noting that adaptive plasticity in Hawaiian Euphorbia specifically may have been furthered by their allopolyploid origin resulting in increased heterozygosity (Yang and Berry, 2011).

Evolution of the tree habit in C 4 Euphorbia
There are many factors that drive, and constrain, the evolution of the tree habit. These include, but are not limited to, protection from animal herbivory, improved dispersal, avoiding (self-) shading, and maintaining water balance. It is unclear what has driven the evolution of trees in Chamaesyce, as Hawaiian Euphorbia tree taxa do not appear to perform better than closely related shrub taxa by measure of abundance.
In terms of evolutionary constraints, in order to diversify to a forest understorey niche, such as that of E. herbstii, a C 4 tree must acquire some shade tolerance. However, shade tolerance does not universally constrain the evolution of the tree habit: C 4 shrubs can displace grasses in high-light scrubland, as is observed for the C 4 shrubs Atriplex confertifolia and A. canescens (Sage and Sultmanis, 2016), and the C 4 tree Euphorbia olowaluana is a pioneer species on newly formed Hawaiian lava fields, occurring sparsely in high-light conditions (see fig. 1H in Yang and Berry, 2011). Therefore, the ability to develop shade-tolerant leaves alone does not dictate whether or not a C 4 plant can evolve the tree habit, and there may be other factors acting to constrain the evolution of trees in existing C 4 lineages, such as the age of these lineages (Sage and Sultmanis, 2016).
The evolution of the tree habit in a herbaceous C 4 ancestor requires sufficient evolutionary time following the appearance of the C 4 trait. C 4 trees, and C 4 shrubs that become arborescent with age, are found in two of the oldest C 4 eudicot lineages: Euphorbia (19.3 Mya) and tribe Salsoleae sensu stricto (Chenopodiaceae, 23.4 Mya), respectively (Table 1; Sage and Sultmanis, 2016). However, in the case of Hawaiian Euphorbia, 19.3 My is much longer than the time required for the transition from herbaceous to tree lifeform: the initial colonization event of the Hawaiian Islands was ~5 Mya and the true tree species themselves are ~1 My old (Yang et al., 2018). Therefore, it may be more accurate to say that it is not evolutionary time, but the rate of evolution that can act to constrain the transition to a tree lifeform in a C 4 lineage. Many eudicot families that have C 4 lineages also have trees, but it may be that the C 4 state limits the rate at which the tree lifeform can be acquired within C 4 lineages by reducing adaptive plasticity (Sage and McKown, 2006;Bellasio and Lundgren, 2016). The aforementioned high adaptive plasticity of Hawaiian Euphorbia, or the fact that the lineage had woody ancestors that had previously undergone a transition to the herbaceous state, may have favoured a comparatively rapid evolution of the tree lifeform in this C 4 lineage (Horn et al., 2014). Interestingly, the ancestor of the C 4 lineage in Salsoleae may have been a shrub or sub-shrub (Schüssler et al., 2017), so while the transition to true tree has not been completed and thus is slower than that in Euphorbia, they may have been advantaged in this transition by acquiring C 4 photosynthesis in an already woody or semi-woody ancestor.
Passive symplastic phloem loading and/or hydraulic limitation negate the benefits of C 4 anatomy in trees Tree height, and indeed the tree growth form, is limited by difficulties in sustaining water and sugar transport over the long pathlength (Ryan and Asao, 2014;Liesche et al., 2017;Savage et al., 2017). We propose that three elements of sugar and water transport design can contribute to limitations on the evolution of C 4 trees.
First, trees tend to exhibit high numbers of plasmodesmatal connections between mesophyll cells and minor vein cells that are devoted to phloem loading, a phenotype that is frequently associated with a passive symplastic loading mechanism (Davidson et al., 2011). The persistently strong sugar sinks of trees, which have meristems and storage sites in their trunks and roots, and also high rates of photosynthesis and thus photosynthetic export from leaves, may actually select for passive symplastic phloem loading as it is less energetically demanding than active mechanisms (Turgeon, 2010). C 4 species, on the other hand, have a high density of plasmodesmatal contacts between mesophyll and bundle sheath cells to allow for the flux of intermediate metabolites (Sowiński et al., 2008;Danila et al., 2016). In a C 4 passive phloem loader, the export of photosynthate would require a large number of plasmodesmatal contacts between bundle sheath and phloem cells. However, this would form a complete plasmodesmatal route from mesophyll to phloem, with C 4 intermediates moving from mesophyll to bundle sheath (and back), while sugar loading proceeds from the bundle sheath into the phloem. Due to the passive nature of this process and lack of compartmentalization, it would be difficult to regulate the flux of C 4 metabolites; that is, prevent leakage of C 4 intermediates into the phloem. Avoiding leakage of C 4 sugars may place a limit on phloem loading via plasmodesmata and thus on passive loading.
Secondly, the combination of C 4 intermediate diffusion with direct plasmodesmatal pathways for sugar transport from the mesophyll to the bundle sheath and into the phloem might be very difficult to sustain, given that sugar movement would be against the transpiration stream. Theoretical analysis has shown that transpiration-induced bulk flow from veins to stomata and passive sugar loading into the phloem by diffusion can co-exist (Rockwell et al., 2018), but these analyses assumed very low plasmodesmatal fluxes between the bundle sheath and mesophyll, as is typical of C 3 species (i.e. without the extensive plasmodesmatal contact typical of C 4 species). More work is needed to determine if transpirational counterflow might present an obstacle to a C 4 passive loader.
Thirdly, taller trees tend to have upper canopy leaves with both lower leaf water potentials and a greater heterogeneity in cell water potentials, due to the tension associated with gravity, greater resistance pathlengths, and exposure of canopies to strong fluctuations in light and temperature (Zwieniecki et al., 2004;Burgess and Dawson, 2007). Plasmodesmata may lose transport capacity when strong pressure differences are generated between adjacent cells or tissues (Oparka and Prior, 1992). Thus, the C 4 pathway, depending on plasmodesmatal transport, may not be feasible at very negative leaf water potentials and/ or given the large heterogeneity of water status within the leaf, and so its evolution may be precluded in trees, particularly tall canopy-forming species unlike Hawaiian Euphorbia (Table 1).
Even if these three limitations could be overcome, the increased photosynthetic efficiency of the C 4 system combined with the reduced ratio of source tissue to phloem requires an increase in phloem loading efficiency to avoid accumulation of photosynthate in the leaves. In C 4 grasses, this is achieved by two mechanisms: first, by up-regulation of active transporters for bundle sheath sugar export, such as maize SWEET-13, a transporter that was duplicated and retained during C 4 evolution (Emms et al., 2016); and secondly, by increasing plasmodesmatal density at the interface of the bundle sheath and the vascular parenchyma to increase passive photosynthate transport in plants grown under high-light conditions (Sowiński et al., 2007). Such adjustments would probably not be possible or effective in a C 4 tree, in the first instance owing to the absence of an active loading mechanism, and in the second instance due to the aforementioned limitations on plasmodesmatal transport. Inability to increase the maximum rate of phloem loading and associated accumulation of leaf non-structural carbohydrates would result in a downward adjustment of photosynthetic capacity that could partially or fully negate the potential benefits of the C 4 system in a tree (Paul and Foyer, 2001;Paul and Pellny, 2003).
How the trees of Hawaiian Euphorbia have circumvented these limitations is unknown, and little is known of their phloem loading mechanism. Notably, the anatomical diversity of Hawaiian Euphorbia is exceptional, from their striking disjunct veins, which vary strongly across species, to their variation in growth form, leaf surfaces, and leaf cross-sectional anatomy (Herbst, 1971;Koutnik, 1987;Horn et al., 2012). More work is needed to discover how phloem loading relates to this diversity, and if a specialized anatomy evolved that mitigates limitations on C 4 tree evolution dictated by phloem processes.

Conclusions
Previous commentary on the rarity of C 4 trees has pointed towards hypotheses of physiology and life history, and found that there is no single explanation that is satisfactory, with Hawaiian Euphorbia (and possibly Salsoleae sensu stricto, Table 1) acting as exceptions to every argument. Each hypothesis seems to have a caveat, whereby if a species is exceptionally shade tolerant, is markedly efficient at sunfleck use, utilizes a particular C 4 biochemical subtype, is notably morphologically diverse, evolved in an especially low competition environment, or has circumvented difficulties in phloem loading, then it may be the exception to the rule. Only by examining all of the interconnected aspects of a complex trait such as C 4 photosynthesis can we begin to understand why a few unique trees have travelled the pathway to C 4 , whereas other would-be C 4 trees cannot complete the journey. New directions into understanding the rarity of C 4 trees also merit investigation, including comparative genomic approaches and investigations into any role that the characteristic latex and laticifers of Euphorbia plants may play in overcoming limitations imposed by passive symplastic phloem loading and hydraulic constraints. These Hawaiian Euphorbia represent a crucial resource in advancing our understanding in these areas. However, they are becoming increasingly threatened in their native habitat: 10 taxa are state and federally listed as endangered and several others are observed to be rare (MJS-K, unpublished data). The narrow endemism of these species and lack of appropriate, protected habitats for conservation mean they are vulnerable to fires, invasive species, human activity, and climate change.
With Hawaiian Euphorbia most probably arising from a herbaceous-to-woody transition in a C 4 ancestor (Yang and Berry, 2011), it is still unclear whether an existing tree could evolve the C 4 trait, and there are currently no known C 3 -C 4 intermediate tree species to indicate that this is a possibility. Indeed, the low quantum yield of these C 3 -C 4 intermediates could mean that, at least under an existing canopy, the transition to C 4 would require the traversal of an adaptive trough (Monson et al., 1986;Stata et al., 2019). Additionally, the occurrence of passive symplastic phloem loading could also act as a barrier to C 4 evolution in existing trees. While these considerations may not apply to the evolution of a tree habit in a C 4 ancestor, this alternative evolutionary path is still constrained by adaptive plasticity (Sage and McKown, 2006;Bellasio and Lundgren, 2016). Thus, both routes to the evolution of a C 4 tree are potentially tortuous, which may together explain the global rarity of C 4 tree species.