Unified changes in cell size permit coordinated leaf evolution


Author for correspondence:

Tim Brodribb

Tel: + 61 362261707

Email: timothyb@utas.edu.au


  • The processes by which the functions of interdependent tissues are coordinated as lineages diversify are poorly understood.
  • Here, we examine evolutionary coordination of vascular, epidermal and cortical leaf tissues in the anatomically, ecologically and morphologically diverse woody plant family Proteaceae.
  • We found that, across the phylogenetic range of Proteaceae, the sizes of guard, epidermal, palisade and xylem cells were positively correlated with each other but negatively associated with vein and stomatal densities. The link between venation and stomata resulted in a highly efficient match between potential maximum water loss (determined by stomatal conductance) and the leaf vascular system's capacity to replace that water. This important linkage is likely to be driven by stomatal size, because spatial limits in the packing of stomata onto the leaf surface apparently constrain the maximum size and density of stomata.
  • We conclude that unified evolutionary changes in cell sizes of independent tissues, possibly mediated by changes in genome size, provide a means of substantially modifying leaf function while maintaining important functional links between leaf tissues. Our data also imply the presence of alternative evolutionary strategies involving cellular miniaturization during radiation into closed forest, and cell size increase in open habitats.


Recent characterization of genes and core regulatory networks has revolutionised our understanding of how tissues develop. However, the development of individual tissues is only one requirement for building complex organisms. Another, less understood process is how the development of spatially discrete but functionally interdependent tissues is coordinated. One possible mechanism for such coordination is colocation of primordial tissues. Thus, lymphatic and blood-carrying vessels of mammals develop from a common embryonic vascular system, and the xylem and phloem of plants derive from a shared cambium. However, complex organisms also depend on the coordinated development of many tissues with different origins (Cavalier-Smith, 2005); for example, lung capacity, vascular volume and muscle mass are necessarily coordinated (Rubner, 1883). Similarly, developmental coordination is essential for plants because their primordial tissues have indeterminate growth. Thus, plants can show great plasticity in response to the environment, but this plasticity is only effective if the diverse tissues involved remain functionally coordinated.

One important example of coordination between discrete tissues is found between the veins and stomata in the leaves of land plants. Branching density in the leaf vein network determines water transport efficiency of the lamina (leaf hydraulic conductance), which is closely linked to maximum rates of photosynthesis (Brodribb et al., 2005) and transpiration (Boyce et al., 2009). Because leaf vascular networks replace water lost by evaporation during the uptake of CO2 for photosynthesis (Sack & Holbrook, 2006), plants with higher rates of photosynthesis per unit leaf area lose more water (Cowan & Farquhar, 1977) and thus demand greater investment in leaf veins (McKown et al., 2010). This investment comes largely as increased branching of leaf minor veins, because a greater density of minor veins delivers water closer to sites of evaporation in the leaf (Brodribb et al., 2007), leading to increased transport efficiency (Sack & Frole, 2006). However, these veins are expensive to synthesize, and plants are likely to coordinate the production of photosynthetic and water supply tissues to maximize returns on investments in the water transport system (Brodribb & Jordan, 2011). Furthermore, while vein density determines water supply in the leaf, the density of stomata determines maximum rates of water loss and photosynthesis, and thus maintaining a balance between these traits during adaptation to the environment should be of high functional and adaptive importance. Such coordination has been demonstrated both within trees during plastic adaptation to light (Murphy et al., 2012) and between species (Edwards, 2006; Dunbar-Co et al., 2009; Zhang et al., 2012).

However, little is known about how this critically important link between vascular and stomatal tissues is maintained. A recent study of a tree species showed that plasticity in epidermal cell size changed vein and stomatal density in concert during light acclimation. Hence, larger epidermal cells in the shade result in larger leaves that have lower densities of veins and stomata than sun leaves (Murphy et al., 2012). This coordinating role of cell size during plastic adaptation of leaves to different evaporative and photosynthetic conditions of sun and shade raises the prospect that changing cell size could also be an important mechanism for evolutionary adaptation in plants.

A correlation between cell volume and genome size has been long recognized as a fundamental feature of eukaryotic organisms (Mirsky & Ris, 1951; Cavalier-Smith, 1985); however, the evolutionary significance of variation in cell size, and associated genome size in plants and animals, has been hotly debated (Cavalier-Smith, 1978, 2005; Petrov, 2001; Hodgson et al., 2010). In animals, transitions in cell and genome size are implicated in several important evolutionary transitions (such as the evolution of birds from dinosaurs; Organ et al., 2007), but in plants the adaptive significance of cell size variation remains obscure. Attempts to account for the enormous range in genome and cell size in plants have recently focused on variation in stomatal size as a potentially important functional consequence of variable cellular and nuclear volume (Beaulieu et al., 2008). Theory and observation suggest that large stomata are associated with low rates of gas exchange as a result of limits on the packing density of guard cells (if stomata become larger, then fewer can fit on the leaf surface), and diminishing benefits in terms of maximum diffusive conductance of larger, deeper pores (Franks & Beerling, 2009). Other potentially important tissues that share size-constrained functional properties include leaf veins, which have analogous associations between the size of cells and the density (Field & Brodribb, 2013) and conductivity (Sack & Frole, 2006; Brodribb et al., 2007) of the vascular system. Epidermal cell size also appears to be a primary determinant of the final size of leaves, as well as influencing the thickness of the photosynthetic mesophyll (Pérez-Pérez et al., 2011). Here we examine how these interconnected systems in the leaf respond to family-wide variation in cell size.

Considering the diversity of influences that cell size has on leaf physiology, we investigate how key functional attributes of leaves remain coordinated if cell size changes. This is of particular significance considering that cell size and genome size appear not only to be rather labile within angiosperms, but also to exhibit some long-term adaptive patterns (Masterson, 1994; Franks et al., 2012b). In this study, we examine the relationship between cell size and functional anatomy in the leaves of a morphologically and ecologically diverse sample of Proteaceae trees and shrubs. The primary question we address in this study is whether the cell sizes in functionally linked tissues are coordinated in such a way as to preserve integrated function. Specifically we hypothesize that greater vein density and stomatal density should be associated with smaller cell size, leading to coordination in tissues related to water supply and water loss in the leaf.

Materials and Methods

The family Proteaceae was chosen for three reasons: it is a morphologically and ecologically diverse group that has a well-documented history of evolutionary adaptation in response to changing climate over the Cenozoic period (Jordan et al., 2008; Sauquet et al., 2009); it has a well-studied phylogeny, thus allowing phylogenetically independent analysis of relationships; and it is an ecologically important group in the southern hemisphere where species range from trees in tropical rainforest to shrubs in the arid zone. We sampled cell size and densities from 48 species and stomatal size from 417 species of Proteaceae from all major branches of the phylogeny.

Species were categorized as being from open vegetation or closed forest according to descriptions from regional floras. Closed canopies are typically > 70% canopy cover, which is generally only achieved in rainforest communities. Proteaceae species are typically canopy species, so regardless of habitat type, all leaves were collected in the field from sun-exposed branches. In most cases, leaves were sampled from three trees and immediately fixed in FAA (50% ethanol, 5% (v/v) acetic acid and 3.7% (v/v) formaldehyde). Leaves were returned to the laboratory where they were soaked in water in preparation for anatomical sectioning. The leaf area and mass of at least 10 leaves per species were measured to yield leaf mass per unit area (LMA).

Stomata and vein density

Paradermal sections of leaves were made using a handheld razor blade to remove the adaxial epidermis and palisade, exposing the minor veins. Sections were then bleached in commercial household bleach (50 g l−1 sodium hypochlorite and 13 g l−1 sodium hydroxide) until clear. Bleach was removed by washing, and sections stained in 1% toluidine blue for 30 s to colour the lignin-rich veins. Finally sections were mounted in phenol glycerine jelly and photographed with a Nikon Digital Sight DS-L1 camera (Melville, NY, USA) mounted on a Leica DM 1000 microscope (Nussloch, Germany) with a ×10 objective. ImageJ (http://rsbweb.nih.gov/ij/index.html) was used to measure the total length of venation in five fields of view that were aligned midway between the midrib and the margin. Wire frames of the veins were drawn manually and their total length counted.

Stomatal density was measured either directly from the paradermal sections, or, where this was not possible, from stomata-bearing cuticles prepared from the same leaves on which vein density was measured. The cuticles were prepared by soaking leaf samples in warm 10% aqueous Cr2O3 until clear, rinsing thoroughly, staining with dilute (< 0.1%) crystal violet, rinsing, cleaning with a single-hair paintbrush (if necessary), and then mounting on microscope slides, in phenol glycerin jelly. Stomatal densities were measured from digital photomicrographs of the cuticle preparation at ×50 magnification using the counting tool in ImageJ. At least three fields of view were measured from each leaf section.

Cell sizes

Stomatal size was determined from photographs of the prepared cuticles described in the previous section at ×40 magnification. Guard cell length and width were measured on 20 stomata from each stomata-bearing surface of the leaves.

Sizes of epidermal cells were measured from paradermal sections of leaves. Length (maximum dimension) and breadth (width perpendicular to the length) were measured on a total of at least 20 epidermal cells from three leaves. Epidermal cell size was then estimated as the square root of the product of length and breadth. The square root was used to ensure that this trait showed the same dimensionality as the other linear dimensions. In one species, Hollandaea sayeriana, epidermal cells could not be reliably differentiated from light microscopy or scanning electron microscopy of paradermal preparations (see, e.g., figs 127 and 128 in Carpenter, 1994). For this species, the size was estimated from mean epidermal cell width in cross-section, using a regression of this dimension against epidermal cell size for the other 47 species.

Palisade and minor vein conduit width were measured in cross-sections prepared from the same leaves as all other anatomical data. Squares of c. 5 mm2 were cut from the region between the midrib and margin and sectioned on a freezing microtome. Sections were stained with toluidine blue and mounted in phenol glycerine jelly. From these sections, we measured leaf thickness and palisade width by taking the maximum cell width from the 20 widest cells per section.

Conduit width was measured in minor veins, identified in the cross-sections by their lack of bundle sheath cells. These minor veins generally contained 10–20 conduits for which lumen width was measured. Three to five minor veins were measured per section.

Maximum stomatal conductance

To examine the balance between water supply determined by vein density, and the demand for water produced by stomatal size and density, we calculated the theoretical maximum leaf conductance (Franks & Farquhar, 2001) to water vapour based on the measured stomatal anatomy using the following equation:

display math(Eqn 1)

where gmax is the maximum leaf stomatal conductance to water vapour (mmol m−2 s−1); d is the diffusivity of water in air (m2 s−1); v is the molar volume of air (m3 mol−1); D is the stomatal density (stomata m−2); a is the maximum pore area (m2); and l is the pore depth (m).

The mean guard cell width was substituted for pore depth based on the assumption that guard cells were approximately circular in cross-section. Maximum pore area was calculated from the guard cell length, assuming that guard cells opened into a circular aperture.

Phylogenetic analysis guard cell length

The phylogenetic distribution of guard cell length was investigated using a phylogeny of 143 species of Proteaceae and six species of its sister clade, Platanus, as an outgroup (Supporting Information, Fig. S1). The phylogeny was created by concatenating the phylogenies of Sauquet et al. (2009), Mast & Givnish (2002) and Mast et al. (2008) with a few additional species added by assuming the monophyly of individual genera. In each of these latter cases, the genera are well defined and accepted. The 149 species were selected to summarize data from 417 of the c. 1700 species of Proteaceae (including all accepted genera) and nine of the 10 species of Platanaceae. Clades in which stomatal density varied by < 15% among species were represented by a single, typical species; in other cases 90th and 10th percentile species were included. The evolution of stomatal length was reconstructed using parsimony with a squared cost assumption, implemented in Mesquite 2.75 (Maddison & Maddison, 2011). Guard cell lengths were allocated to classes based on a log-linear relationship rounded to the nearest μm.

To test for any association between guard cell length and habitat, we used a phylogenetically adjusted one-way ANOVA comparing open vegetation species and closed forest species. This was using a phylogenetic generalized linear model based on this phylogeny, assuming branch lengths of 1, and implemented in the ‘ape’ package of R (Paradis et al., 2001).

Phylogenetically independent correlations

To assess the strength of association between pairs of traits, Pearson product–moment correlation coefficients and phylogenetically adjusted correlation coefficients were calculated. Because stomatal density and vein density were expected to show curvilinear associations with other dimensions, all of the correlations involving these two traits were on a log-log basis. The other correlations were all linear-linear. Pairwise scatter plots confirmed the linear nature of all these relationships. The phylogenetically adjusted correlations were performed using phylogenetically independent contrasts generated using the ‘ape’ package of R (Paradis et al., 2001), based on the phylogeny described earlier but cut down to the 48 species for which the full complement of traits were available.


Interspecies variation

In our sample of 48 species of Proteaceae, we found enormous interspecific diversity in all aspects of leaf morphology and anatomy (Fig. 1). Physical attributes of leaves such as lamina size varied by up to three orders of magnitude (mean leaf sizes ranging from < 1 to > 500 cm2 in the largest species), and leaf thickness by more than fourfold (Table 1). Very large differences between species were observed in the densities of minor veins (ranging from 2.3 to 13.2 mm mm−2) and stomata (from 44 to 520 mm−2). Cell size also varied substantially between species, with a fourfold range in the widths of epidermal and palisade cells and the length of stomatal guard cells (Table 1).

Table 1. Characteristics of species used in this study
SpeciesStomatal distributionLamina thickness (mm)Epidermal cell size (μm)Palisade cell width (μm)Guard cell length (μm)Diameter of vessel lumens (μm)Vein density (mm mm−2)Leaf mass per unit area (g m−2)Stomatal density (mm−2)Leaf area (cm2)
  1. Stomatal distribution is either amphi- or hypostomatic. Stomatal densities are for the abaxial leaf surface, and adaxial densities are given in parentheses for amphistomatic species.

Conospermum longifolium Amphistomatic0.5932.923.125.24.977.95174150 (143)8.7
Isopogon fletcheri Amphistomatic0.59447. (90)11.3
Leucadendron pubescens Amphistomatic0.56139.120.941.75.644.82198130 (99)0.93
Leucospermum cordifolium Amphistomatic0.6146.228.940.511.234.62255103 (99)8.3
Persoonia lanceolata Amphistomatic0.40592.425.379.78.864.0313427 (23)5.2
Persoonia muelleri Amphistomatic0.71464.448.267.76.432.2926927 (27)1.9
Protea cynaroides Amphistomatic0.58763. (3)44.5
Protea nitida Amphistomatic0.51531.922.338.98.556.60288141 (124)32.1
Protea repens Amphistomatic0.52530.621.3616.259.5223060 (56)5.0
Synaphea petiolaris Amphistomatic0.39650.421.545.98.945.9116948 (47)14.5
Synaphea spinulosa Amphistomatic0.39346.122.7464.495.6824473 (67)24.6
Agastachys odorata Hypostomatic0.72439.740.4605.383.40273463.8
Alloxylon pinnatum Hypostomatic0.31331.520.739.37.424.83140989.8
Athertonia diversifolia Hypostomatic0.17618.822.120.84.437.1394267660
Austromuellera trinervia Hypostomatic0.19616.714.7273.3812.80150521429
Bankia grandis Hypostomatic0.34522.
Bellendena montana Hypostomatic0.63831.832.842.74.564.972291281.2
Brabejum stellatifolium Hypostomatic0.33621.025.527.78.0010.6418527918.3
Buckinghamia celsissima Hypostomatic0.28919.014.823.33.658.5115631239.2
Cardwellia sublimis Hypostomatic0.22221.
Carnarvonia araliifolia Hypostomatic0.36240.822.2394.706.33137145400
Catalepidia heyana Hypostomatic0.30720.
Cenarrhenes nitida Hypostomatic0.62927.830.433.53.914.672262107.5
Darlingia ferruginea Hypostomatic0.24431.713.538.65.7010.8023417799.4
Eidothea hardeniana Hypostomatic0.2528.120.9295.788.2820412919.7
Embothrium coccineum Hypostomatic0.25429.918.930.75.076.8813716117.8
Floydia praealta Hypostomatic0.24616.313.326.34.1912.4411032658.8
Gevuina avellana Hypostomatic0.34927.723.
Grevillea hilliana Hypostomatic0.33821.817.226.14.877.6216324667.9
Helicia australasica Hypostomatic0.24122.018.7304.808.1311423128.2
Hicksbeachia pilosa Hypostomatic0.2739.641.727.44.426.25151106560
Hollandaea sayeriana Hypostomatic0.39824.139.926.55.915.21202189123
Lambertia inermis Hypostomatic0.24434.019.741.14.1310.392611250.78
Lasjia whelanii Hypostomatic0.33819.820.725.63.9811.4610935919.1
Lomatia tinctoria Hypostomatic0.32531.121.8284.204.8324416110.1
Megahertzia amplexicaulis Hypostomatic0.29117.319.124.43.9712.79114303150
Musgravea heterophylla Hypostomatic0.21817.318.623.14.2813.21147444259
Neorites kevedianus Hypostomatic0.22124.618.4323.896.6383196368
Opisthiolepis heterophylla Hypostomatic0.2720.814.520.54.415.8889250435
Orites diversifolius Hypostomatic0.61216.522.
Orites milliganii Hypostomatic0.6220.820.850.34.375.30300792.6
Placospermum coriaceum Hypostomatic0.61185.730.762.78.753.1811143249
Roupala pseudocordata Hypostomatic0.4516.916.837.36.883.7722513156
Sphalmium racemosum Hypostomatic0.45268.027.859.611.013.521914513.9
Stenocarpus sinuatus Hypostomatic0.32922.416.831.35.896.34202196155
Telopea truncata Hypostomatic0.64127.620.727.84.674.432261294.9
Triunia montana Hypostomatic0.26931.623.733.74.336.669213730.5
Xylomelum pyriforme Hypostomatic0.35925.
Figure 1.

Representative images from two sampled Proteaceae species with contrasting leaf anatomy. Images show, from left to right, stomatal size and density; vein density; and cross-sections illustrating epidermal and palisade cell sizes. All images are at the same magnification to highlight the very large differences in the sizes of cells and densities of veins and stomata.

Stomata and veins

The densities of stomata (Dstomata) and minor veins (Dvein) were very strongly linearly correlated among all species with an almost proportional relationship encompassing all species (Dstomata = 23.95 Dvein + 25; < 0.0001; Fig. 2). The species with the lowest Dstomata (44 stomata mm−2) also had the lowest vein density (2.3 mm mm−2), while the highest Dstomata (521 stomata mm−2) was found in a rainforest species with a high vein density of 12.8 mm mm−2. The density of veins showed no relationship with leaf area, and the density of stomata showed a weak positive relationship with leaf area; however, both densities showed strong and highly significant negative correlations with leaf thickness (Table 2).

Table 2. Correlations among pairs of traits
 Guard cell lengthStomatal densityVein densityEpidermal cell widthXylem lumen diameterPalisade cell widthLamina thicknessLeaf mass per unit areaLeaf area
  1. Numbers above the diagonal are unadjusted Pearson correlation coefficients, numbers below the diagonal are phylogenetically adjusted coefficients.

  2. ***, P < 0.001; **, P < 0.01; *, P < 0.05; ns, P > 0.005.

Guard cell length −0.91 ***−0.59 ***0.84 ***0.56 ***0.47 ***0.56 ***0.27 ns−0.30 *
Stomatal density−0.89 *** 0.71 ***−0.88 ***−0.63 ***−0.58 ***−0.58 ***−0.29 *0.41 **
Vein density−0.62 ***0.72 *** −0.65 ***−0.49 ***−0.65 ***−0.68 ***−0.29 *0.34 *
Epidermal cell width−0.79 ***−0.86 ***−0.72 *** 0.64 ***0.51 ***0.43 **0.04 ns−0.13 ns
Xylem lumen diameter−0.44 **−0.52 ***−0.42 *0.53 *** 0.28 ns0.35 *0.18 ns−0.24 ns
Palisade cell width−0.40 *−0.58 ***−0.69 ***0.57 ***0.30 * 0.58 ***0.28 ns−0.05 ns
Lamina thickness−0.58 ***−0.53 ***−0.65 ***0.41 **0.36 *0.59 *** 0.62 ***−0.44 **
Leaf mass per unit area0.35 *−0.30 *−0.35 *0.09 ns0.31 *0.30 *0.65 *** −0.48 ***
Leaf area−0.34 *0.38**0.38 **−0.07 ns−0.21 ns−0.12 ns−0.54 ***−0.57 *** 
Figure 2.

Correlation between the mean density of stomata on both leaf surfaces and the mean density of veins in the leaves of 48 species of Proteaceae trees and shrubs. Despite an enormous range in leaf size, shape and habitat, a highly significant linear correlation (r2 = 0.45; < 0.0001) between the densities of water supply tissue and stomatal pores was observed. See Table 1 for species list.

Cell size

The sizes of stomatal, epidermal, xylem and palisade cells were all interrelated. The most important associations were found between epidermal cell width and both the guard cell length (Table 2, Fig. 3a) and xylem lumen diameter in minor veins (Table 2), all of which were related by very strong linear correlations among species. Palisade cell width was also significantly correlated with these other cell types, but these correlations were not as strong (Table 2).

Figure 3.

All species showed coordinated changes in species mean cell sizes from stomatal, epidermal, vein and palisade tissues. Linear regressions were highly significant in all cases regardless of whether data were phylogenetically adjusted (Table 2) or unadjusted (shown here). See Table 1 for species list.

Cell sizes were significantly linked to key functional traits, Dvein, Dstomata and the thickness of leaves, but not leaf area (Table 2). The strongest association was between the size of stomatal guard cells and Dstomata, with diminishing guard cell length in species with higher Dstomata (Lguardcell = 255Dstomata−0.43; r2 = 0.82). Confirming that this relationship is probably driven by spatial constraints, the Dstomata in amphistomatic species fitted the same regression against stomatal size as hypostomatic species only when just stomata on the lower leaf surface were considered (Fig. 4). Other cell sizes, particularly xylem and epidermal cell size, were also strongly correlated with Dstomata (Table 2). Similar to Dstomata, we found that Dvein was strongly correlated with cell size, with highly significant correlations (< 0.001) among species between Dvein and all cell types (Table 2). High vein density was strongly associated with small cell size in stomata, epidermis, palisade and xylem cells across all species.

Figure 4.

Spatial limits to stomatal packing produce a general relationship between the size and density of stomata on the lower leaf surface (black circles) in all Proteaceae. Larger stomatal size leads to lower density, following a predictable relationship; stomatal length = 289.1(stomatal density)−0.43. Species with stomata on both leaf surfaces (amphistomy; green circles) were able to achieve higher total stomatal densities (the sum of both leaf surfaces) by avoiding the strict spatial limits associated with confining stomata to the lower leaf surface. See Table 1 for species list.

Functional coordination

Maximum predicted stomatal conductances to water vapour were found to be strongly correlated with Dvein (Fig. 5). This pattern among species conformed to the predicted linear relationship between the capacity to supply water (proportional to Dvein) and the capacity of leaves to lose water (proportional to maximum stomatal conductance). Species with stomata on both leaf surfaces were not expected to fall on the same relationship (see the Materials and methods section) and, indeed, we found a nonsignificant slope in these 11 species (Fig. 5).

Figure 5.

Coordination in cell size leads to a strong correlation between vein density and maximum stomatal conductance. In all species with stomata confined to the lower surface (black circles), the density of veins providing water supply to the leaf (a proxy for the water transport efficiency of the leaf; Brodribb et al., 2007) was linearly related to the maximum potential rate of water loss, modelled from the stomatal anatomy and density. Leaves with stomata distributed on both surfaces (green circles) did not conform to the general relationship found in typical hypostomatic species and were able to produce higher maximum stomatal conductances from the same venous supply.

Phylogenetic patterns in stomatal size

Stomatal size data spanning the full generic diversity of Proteacae demonstrated different patterns in different clades (Fig. 6). Moderately large stomata (31–54 μm) were found almost throughout the large, and mainly open-habitat, subfamily Proteoideae, with the only major variation being the appearance of very large stomata within the recently evolved genus, Protea (also of open habitats), and smaller stomata in a few genera. By contrast, the subfamily Grevilleoideae, in which most of the phylogenetic diversity is in rainforest, mostly has small stomata (< 31 μm), particularly in rainforest genera. Finally, the mainly open-habitat subfamily Persoonioideae shows markedly large stomata (> 60 μm). The ancestral state reconstruction suggests the independent evolution of both very large stomata (in Persoonioideae, Protea, Agastachys and Strangea) and small stomata (in several clades within Grevilleoideae and within a few genera of Proteoideae) (Fig. S1). The emergent pattern of small stomata in rainforest clades and large stomata in open-habitat clades is confirmed by the presence of a phylogenetically independent association between habitat type and stomatal size, with clades of open environments being very significantly more likely (< 0.001) to have large stomata than closed forest clades.

Figure 6.

Stomatal length and habitat type (open vegetation or closed forest, or both) across the full phylogenetic range of Proteaceae, with the evolution of stomatal length reconstructed using parsimony. A phylogenetically adjusted one-way ANOVA showed significantly greater stomatal length in open vegetation species than in closed forest species (< 0.001). Persoonioideae has been abbreviated to Pers. See Supporting Information Fig. S1 for individual species labels.


As leaf anatomy and morphology evolve and adapt in response to environmental change, there is the potential for functionally interdependent tissues to change independently of one another. Without some coordinating mechanism, this would lead to inefficiency or plant death. Our results show that coordinated changes in cell size can function as a linkage between the vascular system and epidermis during adaptive variation in leaves of Proteaceae. Over an extreme range of leaf size, morphology and habitat, our sample of 48 species of shrubs and trees showed strongly correlated changes in the sizes of distinct cell types (stomatal, epidermal, xylem and palisade cells) in the leaf. This resulted in a unified relationship between the density of stomata on the leaf surface and the density of vein branching in the lamina. As these tissues are responsible for regulating interlocked processes of water delivery and water loss in leaves, modification of cell size therefore provides a rapid way for plants to adapt and evolve to the prevailing conditions without compromising the coordination of component tissues necessary for the whole leaf to function effectively (Fig. 7).

Figure 7.

Schematic representation of the linkages between variation in cell size and species adaptation, explaining why cell size should be coordinated and why variation in cell size is a useful adaptive tool. As long as mutations in cell size produce parallel changes in the size of all cell lines in the leaf, the critical link between water supply and transpiration/photosynthesis will be maintained. Selection can then act upon the functional outputs of cell size variation in the leaf, which are photosynthetic performance and leaf thickness, both of which are highly adaptive under different conditions. The selective advantage of high rates of photosynthesis is relatively straightforward, while leaf thickness is involved in a suite of associated functional characters, such as leaf mass per unit area (LMA), stomatal positioning and leaf water content (see the Discussion section).

Variation in the sizes of cells and genomes appears to be closely related across biological kingdoms (Cavalier-Smith, 2005), but the selective processes driving this variation has been the subject of considerable controversy. Based upon a linear correlation (< 0.01; data not shown) between the stomatal size data presented here for Proteaceae and genome size published by Stace et al. (1998), we can say that Proteaceae provide an exemplary demonstration of this so called ‘C-size paradox’. A commonly cited explanation for cell and genome size variation suggests that different cell sizes may be suited to different ecological strategies, and in particular whether a species is the product of r- or K-type selection (Cavalier-Smith, 1978). Owing to an apparent negative correlation between cell size and rates of cell division (Van't Hof & Sparrow, 1963), large cells/genomes are proposed to result from K-selection, yielding slow-growing species, while small cells are suggested as being associated with r-selected species, with rapid turnover. This idea has received some support from studies in vertebrates (Vinogradov, 1995), and has been tested in seed plants on the premise that K-selected species are typically those with high LMA and hence that LMA should be positively correlated with genome size. Although early studies provided some support for this hypothesis (Knight et al., 2005), a larger study found nonsignificant associations within a phylogenetically independent framework (Beaulieu et al., 2007), and similarly within the Proteaceae we found no correlations between cell size and LMA (Table 2).

Rather than expecting cell size to affect leaf economics in a predictable way, we argue here that by coordinating cell sizes in the leaf, plants are free to adapt the water transport and photosynthetic gas-exchange properties of leaves while still maintaining an optimal balance between water transport and water loss tissues. Coordination in stomatal and hydraulic physiology provides important functional advantages in terms of maintaining optimal investment and safety (Brodribb, 2009) and this pressure may be the adaptive driver that has led to cell size-dependent coordination between xylem and stomatal tissues in the leaf. Along with other studies showing correlations between cell size and the density of stomata and veins (Murphy et al., 2012; Zhang et al., 2012), the data here from such a diverse family as the Proteaceae provide strong evidence of the primacy of the stomata–vein linkage. Size correlations between other leaf cells, such as epidermal cells and palisade cells (Fig. 3), may have functional significance as well, but it is also possible that the correlated size of these cells is merely a consequence of a common genetic control of cell sizes, perhaps via genome size.

Other studies have indicated the importance of cell size in regulating the density of stomata on the leaf, because constraints on packing density mean that fewer large stomata can fit on the leaf than smaller stomata (Beaulieu et al., 2008; Franks & Beerling, 2009). Our data show that this relationship is particularly strong among Proteaceae (Fig. 3, Table 2). Owing to the connection between size and density in stomata, a critical implication of reduced stomatal cell size is a dramatic increase in the amount of water and CO2 that can be exchanged over the leaf epidermis, because stomatal density is allowed to increase and pore depth is reduced. The net result of this is that a leaf epidermis built from smaller cells will support a much higher capacity for gas exchange, in terms of both photosynthetic CO2 and transpirational water vapour. However, without a parallel increase in the density of minor veins in the leaf, the increased capacity for photosynthesis cannot be realized because stomata will not be sufficiently irrigated to open fully (Brodribb, 2009). Hence the discovery here, that vein density is also sensitive to cell size in such a way as to change in proportion with stomatal density, is highly significant because the result is a match between water supply and water loss from the leaf. This match is most clearly demonstrated by the proportionality between modelled maximum stomatal conductance and vein density across all species (Fig. 5), indicating that the hydraulic efficiency of the vein network (as determined by vein density; Brodribb et al., 2007) remains in proportion to the water-loss capacity of the leaf. Interestingly, we found that vein density was most strongly correlated with the size of stomatal and epidermal cells, and whilst these cells are the key determinants of stomatal density, they are separated from the xylem and mesophyll cells that make up the vein network and surrounding tissue.

One pathway to evolving enhanced photosynthetic rate in leaves can be easily visualized as a process of selection for small cell size, leading to high densities of stomata and veins in leaves (Fig. 7). Such an evolutionary trajectory appears to have been significant during the evolution of the Proteaceae. Thus, one major clade (the Grevilleoideae) is dominated by taxa with small stomata and high vein densities. This pattern makes ecological sense because cellular miniaturization in this group is associated with species that occupy tropical rainforests, where rapid gas exchange and growth tend to be selected for (Körner, 1994). Interestingly, however, several clades, including two large and ecologically successful groups (Persoonioideae and Protea, each with c. 100 species), have very large stomata. These cases are of particular interest because they imply that apparently less productive, large cell size must also confer selective advantages to plants under some conditions. In the Proteaceae, very large stomata occur almost exclusively in clades of open habitats (Fig. 6), and almost exclusively in species with thick and mostly amphistomous leaves. This association also fits the evidence that selection under high light conditions favours thick leaves and amphistomy (Smith et al., 1998; Cooper & Cass, 2003). Amphistomous leaves are highly efficient in terms of photosynthesis, water use and hydraulic supply (Mott et al., 1982), and the ability to spread stomata more sparsely on both leaf surfaces will greatly ameliorate stomatal crowding constraints, probably allowing stomatal size to increase. It is uncertain whether the primary selective force in open conditions is for thick leaves and hence large cells, ultimately allowing amphistomy to develop, or whether amphistomy is selected for initially, leading to a relaxation in spatial constraints in the leaf and allowing cell size and leaf thickness to increase.

Evidence suggests that amphistomy is strongly associated with high light environments and selected against in shady environments (Carpenter, 1994; Smith et al., 1998). While amphistomy provides for a more efficient use of water (Mott & O'Leary, 1984), mesophyll tissue (Parkhurst, 1978) and vein tissue (because veins can deliver water simultaneously to both surfaces of the leaf), these benefits require substantial illumination of both leaf surfaces, and under low light the significant internal self-shading makes this a highly inefficient architecture. Alternatively, the more common configuration of stomata confined to the lower leaf surface seems to provide plants with a more efficient light-harvesting configuration under shadier conditions. However, the restriction of stomata to the lower leaf surface requires that stomata become densely concentrated and hence vein density must be very high to sustain rapid water transport to this highly evaporative surface. The combination of hypostomy and high rates of gas exchange would therefore produce substantial pressure for simultaneous miniaturization of both stomata and veins (Field & Brodribb, 2013). However, as long as the size of these cells declines in concert, leaves will remain balanced in terms of hydraulic supply and evaporative load. Our data for Proteaceae indicate that this cell size-dependent coordination does operate, and we suggest that this is possibly mediated by the size of the species genome.

Recently, it has been suggested that changes in stomatal and genome size in land plants may be a response to Phanerozoic megacycles of CO2 in the atmosphere, with periods of low CO2 favouring smaller stomata (Franks et al., 2012a). The common link between stomatal and vein density, demonstrated here through cell size, provides an important connection between the evolutionary patterns described for stomatal density and similar patterns reported for vein density (Boyce et al., 2009). Studies of leaf veins across the phylogeny of vascular plants show that, although vein densities range from < 1 to > 25 mm mm−2, only angiosperms are able to produce leaves with vein densities in the upper part of this range (> 6 mm mm−2). This unique capacity in angiosperms has been discussed as an important factor contributing to their success over competing plant groups (Brodribb & Feild, 2010), and, similar to the stomatal trend, the rise in vein density requires a decrease in cell size (Field & Brodribb, 2013) and is likely to respond to changes in atmospheric CO2 (Brodribb & Feild, 2010). It seems reasonable to suggest that the transition from ferns and gymnosperms characterized by low densities of veins and stomata to angiosperms with high densities of veins and stomata might be mediated by a decrease in genome size, a proposition supported by the fact that ferns and gymnosperms tend to have large genome sizes relative to angiosperms (Leitch et al., 2005). It remains unresolved, however, whether changing genome size is a driver or a product of macroevolutionary changes in vascular plants. In the context of this question, it is relevant to note that the evolution of high vein density in angiosperms actually occurred after the divergence of early angiosperm lineages, and that the ancestral state for angiosperm vein density is low, despite the fact that the early angiosperms are thought to have small genomes (Masterson, 1994) and therefore potentially small cells. A focus on the evolutionary trajectories of genome size and cell size at the base of the angiosperms will provide important insights into the drivers of this critical evolutionary event, and whether a linkage among genome size, cell size and water use is universal among vascular plant species.