Leaf venation: structure, function, development, evolution, ecology and applications in the past, present and future

1. Dimensions of vascular cells within veins and of whole veins

Plant species vary enormously in the cross-sectional dimensions of xylem, phloem and sclerenchyma cells within leaf veins of given orders (Fig. 2p–s) and in the external diameters of the whole veins (i.e. vascular bundles). This variation contributes to multiple functional differences (Notes S2). All else being equal, larger conduit lumens provide greater xylem flow conductivity and K_leaf, and greater sensitivity to xylem embolism, whereas larger xylem conduit wall thickness relative to lumen diameter provides greater biomechanical strength and resistance to implosion, and longer xylem conduits can contribute to the spread of pathogens across the leaf. Finally, greater phloem conduit diameters can contribute to higher phloem flow conductivities. These differences can drive impacts on leaf gas exchange and its responses to external conditions.

The arrangement of tissues within veins may also influence transport. In the large veins of most angiosperms, the phloem is abaxial to the xylem (Esau, 1977), but species in at least 27 families also have adaxial phloem (Metcalfe & Chalk, 1950). In some species, within the lower-order veins, the vascular cells are separated in discrete bundles, a type of ‘sectoriality’ which may protect the leaf from the spread of embolism within veins (Orians et al., 2005; Brodersen et al., 2012).

The diameters of whole veins can also reflect differences in transport capacity when they contain greater sizes and numbers of xylem and phloem cells (Fisher & Evert, 1982; Russin & Evert, 1985a; Coomes et al., 2008; Taneda & Terashima, 2012). A larger vein diameter provides greater mechanical support and protection for a given vein cross-sectional shape and composition and proportion of lignified cells (Niklas, 1992; Onoda et al., 2011; Méndez-Alonzo et al., 2013), and can better resist animal damage (Read & Stokes, 2006). Investment of mass and volume in petiole and midrib can compensate for weaker lamina to provide whole leaf strength (Méndez-Alonzo et al., 2013). Thicker veins are typical in larger leaves (Sack et al., 2012). In some species large midveins can provide a frame for the lamina to curl, or to allow flexural bending along the midrib, to reduce transpiration and mechanical load (Givnish, 1979; King et al., 1996; Cooley et al., 2004).

The veins entail substantial cost. Across species, the major veins can contribute a substantial amount to the leaf mass per unit area, and their carbon and nitrogen concentrations are correlated with those of the lamina (Niinemets et al., 2007a,b).

2. Vein hierarchy

In the hierarchical leaf vein system of angiosperms, veins of higher branching orders in a given leaf have smaller diameters but greater branching frequencies and lengths (Figs 1a–h, 2j). The bulk of vein length is in the minor veins, whereas substantial vein cross-sectional and projected area are contributed by both major and minor veins, and the bulk of vein volume is in the major veins (Sack et al., 2012). This hierarchy is simplified in earlier-evolved plant lineages (21). Most monocots, including grasses, have a distinct hierarchy of gridded ‘parallel’ or ‘striate’ major veins (Fig. 2e,f), with midribs, and large and intermediate longitudinal veins, analogous to major vein orders, and small longitudinal veins and transverse veins analogous to minor veins (Ueno et al., 2006). Several eudicot lineages have also evolved striate vein systems (e.g. Plantago and phyllodes of Acacia) and certain monocot families have evolved eudicot-like vein systems (e.g. the Araceae and banana families). In palms, the fronds have gridded systems, and the transverse veins can be as large as the large longitudinal veins, and can even be disconnected from some of the longitudinal veins that they cross, suggesting semi-independence within the network (Horn et al., 2009).

As a result of the hierarchy, the diameters of vein segments within a given leaf tend to show skewed frequency distributions when quantified for the whole system, that is, with the majority of segments being of a small size class (Bohn et al., 2002). The hierarchy shows a general scaling across angiosperm species; there are correlations among the diameters of the petiole, midrib and second-order veins (Coomes et al., 2008; Scoffoni et al., 2011; Sack et al., 2012). The hierarchy of external vein dimensions reflects that of the tissues and cells within. From lower to higher vein orders, the xylem and phloem cross-sectional areas decrease with that of the whole vein, as do the numbers, diameters and lengths of xylem conduits, parenchyma and fiber cells, and phloem sieve elements (SEs) and companion cells (CCs), except the diameters of the CCs increase in the minor veins for phloem loading (Plymale & Wylie, 1944; Fisher & Evert, 1982; Russin & Evert, 1985a; Dannenhoffer et al., 1990; Coomes et al., 2008).

The hierarchy of vein dimensions and lengths, and associated shifts in cross-sectional anatomy across vein orders provide multiple benefits (Notes S3). Advantages conferred for water and sugar transport include cost-efficiency; increased contact of higher vein orders with photosynthetic tissues; pressure equilibration across the leaf; and tolerance of damage or blockage in higher-order veins (Zwieniecki et al., 2002; Sack et al., 2004; McKown et al., 2010). Additionally, the vein hierarchy also provides cost-efficiency for biomechanical support and protection against herbivory (Méndez-Alonzo et al., 2013).

3. Vein tapering

In angiosperm leaves, both the first- and second-order veins narrow in diameter along their length (Fig. 1a–h). While in conifers and in some fern pinnae midvein tapering occurs, tapering along the second-order veins is notably absent. Indeed, in many ferns, the second-order (and sometimes first-order) veins widen to supply marginal sori (Fig. 2h,i). Along the tapering veins, the cross-sectional area and number of cells of each tissue type decrease (Fisher & Evert, 1982; Russin & Evert, 1985a; Dannenhoffer et al., 1990).

Tapering provides benefits previously described for venation hierarchy, that is, increased hydraulic capacity relative to construction costs (Zwieniecki et al., 2006; McKown et al., 2010). Vein tapering also provides economical mechanical support, because the load is borne at the base of the petiole and midvein, which operate as a cantilevered beam (Niklas, 1992). Similarly, the loads on second-order veins are mostly borne at the thickened base. Tapering can also allow specialized biomechanical function. In long grass leaves, midvein tapering contributes to the bending of the blade such that the center of the leaf faces sunlight (Moulia & Fournier, 1997).

4. Vein length per unit area (‘vein density’)

The leaf vein length per unit area (VLA; also known as ‘vein density’) varies strongly across species (Fig. 2a–f). In the typical angiosperm, except for those with striate vein systems, VLA is mainly determined by the minor vein length per unit area (minor VLA), which accounts for > 80% of the total vein length (Sack et al., 2012). Across species or genotypes of a given species, high VLA can correspond to a lower number of epidermal and mesophyll cells between veins, and/or to smaller cells (McKown & Dengler, 2007; Manuel Perez-Perez et al., 2011; Smillie et al., 2012).

To quantify VLA, X-ray imaging can be applied for some leaves (Blonder et al., 2012), but for many others, it is necessary to chemically clear the leaf (Sack & Frole, 2006; Scoffoni et al., 2011) or prepare paradermal sections (Brodribb et al., 2007) and microscopically image the minor vein network. Using software, images can be measured manually or automatically, although it is essential to validate automatic methods, and measurement of low-resolution images of venation can underestimate the VLA (see Rolland-Lagan et al., 2009; Price et al., 2011; Dhondt et al., 2012; Sack et al., 2012). The interveinal distance (IVD) has been often measured as a proxy for VLA, as, across species, the two are reciprocally related, although the relationship can vary depending on the topology of the minor veins, and IVD can be much more variable within leaves than VLA (Uhl & Mosbrugger, 1999; Blonder et al., 2011).

A high VLA (Fig. 2b,d,f) has multiple benefits (Notes S4). All else being equal, a high VLA can enable higher K_leaf, greater stomatal density and stomatal conductance and higher rates of gas exchange per leaf area (Sack & Frole, 2006; Brodribb et al., 2007, 2010; Boyce et al., 2009; McKown et al., 2010; Feild et al., 2011; Walls, 2011). By hypothesis, a high VLA can enable greater phloem transport efficiency (Russin & Evert, 1984), but this depends on the distribution of phloem in the vein network, and on species’ loading strategies among other factors. A high VLA may also provide tolerance of fine-scale damage or vein blockage (Raimondo et al., 2003; Sack & Holbrook, 2006; Nardini et al., 2010). Additionally, a high VLA may confer benefits for biomechanics and protection, although, given the low volume of minor veins, this effect would be subtle relative to the influence of major veins and the composition (Choong et al., 1992). In grasses and palm leaves, the small transverse veins reinforce against bending forces (Niklas, 1999). A high VLA may also provide an advantage against herbivory, forcing insects to spend energy to cut through more veins (Vincent 1990; Read & Stokes, 2006). Finally, by contrast, a low VLA (Fig. 2a,c,e) has possible benefits for certain species, for example, species under shade, or high water supply, corresponding to a reduced construction cost and displacement of mesophyll (Sporck & Sack, 2010).

Famously, VLA is often higher for C₄ relative to C₃ species. C₄ species typically have high VLA so that mesophyll cells are at most two cells away from veins to allow diffusion of sugars via plasmodesmata to their enlarged bundle sheath (BS) for decarboxylation and photosynthesis (Ogle, 2003; Sage, 2004; Griffiths et al., 2012). The importance of VLA was clear in the evolution of C₄ within eudicot genera Cleome and Flaveria, as VLA increased from C₃ to C₃/C₄ intermediate and C₄ species (Marshall et al., 2007; McKown & Dengler, 2007). However, the linkage is not entirely necessary; some species perform C₄ within a single cell, and no difference in VLA was found on average between C₃ and C₄ eudicots sampled from sister lineages (Muhaidat et al., 2007). The diffusion of C₄ sugars would be more limited by the number of cells between veins than by distance per se, as transport through plasmodesmatata would be the major resistance to the flux of solute (Haritatos et al., 2000b), and species with larger cells may maintain efficient C₄ despite lower VLA. The consistently higher VLA of C₄ grasses (Ueno et al., 2006) suggests a particular mechanistic basis in that lineage. By contrast, CAM plants, with their typically succulent habit, have low VLA (19).

5. Major vein length per unit area (‘major vein density’)

The major vein length per area (major VLA; Fig. 1a–h) has multiple key functions (Notes S5). All else being equal, a high major VLA can contribute to: higher K_leaf and gas exchange (Sommerville et al., 2012); tolerance of the leaf vein system to disruption by damage, drought, and, potentially, freeze–thaw embolism (Sack & Holbrook, 2006; Scoffoni et al., 2011); equalizing of water potential across the leaf (Zwieniecki et al., 2004b; Ocheltree et al., 2012); and biomechanical support (Lucas et al., 1991; Choong et al., 1992; Niklas, 1999). However, the major veins also involve a substantial construction cost in carbon and nutrient investment (see 3).

These functions of high major VLA would benefit palmate-veined relative to pinnately veined species, that is, those that have multiple first-order veins branching from the base of the leaf (Fig. 1e–h) (Sack et al., 2008). The hydraulic and mechanical protection provided by major vein redundancy would be very important in thinner and larger leaves that are not already protected. Indeed, palmate-veined species occur more commonly in temperate than in tropical floras and are more frequent in large leaves with thin laminae (and low leaf mass per unit area) (Sack et al., 2008; Walls, 2011).

Smaller leaves tend to have higher major VLA intrinsically as a result of their development (14). The drought tolerance and protection conferred by higher major VLA in smaller leaves provides an explanation for the greater abundance of smaller leaves in drier and more exposed habitats (Scoffoni et al., 2011; Fig. 1a–d).

6. Topology: vein ‘types’, loopiness, areolation, and branching angles

Variation in vein topology, that is, the arrangement of network elements, accounts for much of the visually striking diversity of leaf vein systems. Vein systems have been classified with ‘types’ for each order (Hickey, 1973; Ellis et al., 2009). For first-order veins, types include palmate vs pinnate. For second-order veins, types include craspedodromous systems, with the veins running to the margin (Fig. 1e–g), brochidodromous systems with looping second-order veins (Fig. 1a,c,d,h), and eucamptodromous, an intermediate type (Fig. 1b). For third-order veins, types include percurrent, ramified and reticulate.

Some differences in vein types correspond to differences in ‘loopiness’, that is, how often veins of given orders interconnect. Early researchers hypothesized that reticulation (higher-order loopiness) protects against herbivory (Haberlandt, 1914; Wagner, 1979) and, indeed, in leaves lacking reticulation, severing veins kills the distal lamina (Shull, 1934), and models showed that looping hierarchical systems become optimal for transport and distribution only when there are fluctuations in flow, or damage (Bohn & Magnasco, 2007; Corson, 2010; Katifori et al., 2010). An index of loopiness is areolation, with an areole defined as the smallest area of leaf enclosed by veins (Ellis et al., 2009). Across species, number of areoles per unit leaf area correlates with total VLA (Blonder et al., 2011). However, areoles vary in their shape and size within given leaves, some leaves have no areoles (i.e. those with freely ramifying high-order veins), and many species have several tiers of areolation, loops within loops. Recent work has focused on developing other descriptors for loopiness (Katifori & Magnasco, 2012; Mileyko et al., 2012).

Some differences in vein types correspond to differences in VLA for given vein orders. For example, palmate systems have higher first-order VLA than do pinnate systems. Other topologies signify greater connectivity and may provide greater hydraulic conductance and support for a given VLA (Roth-Nebelsick et al., 2001); thus, brochidodromous leaves have greater second-order vein connectivity than do craspedodromous leaves of the same second-order VLA, and have greater leaf mass per unit area and leaf lifespan on average globally (Walls, 2011), probably reflecting their greater abundance among tropical evergreen species (Roth-Nebelsick et al., 2001). Biomechanical support may be provided by the numerous parallel second-order veins in some eudicots (Fig. 2k), as in species of Calophyllum and Micropholis, and the striate system of grasses (Lucas & Pereira, 1990; Onoda et al., 2011).

Vein branching angles are also diverse across species. Across the whole vein system, the angle within each three-way vein junction is proportional to the radii of the connecting veins: smaller veins meet at more acute angles (Bohn & Magnasco, 2007). Mathematical analyses showed this property to be consistent with minimizing flow resistance relative to vein surface area, but also that it could arise during leaf expansion if tensions develop according to the relative strength of the veins and contribute to the final angles (Durand, 2006). In some species, the veins guide the folding of the developing leaf in the bud, and branching angles reflect the manner of folding (Couturier et al., 2012), with implications for bud size and the economics of leaf growth. Leaves with higher second-order vein branching angles can be folded more compactly but require more energy for unfolding (Kobayashi et al., 1998).

7. Free-ending veins

Leaves vary strongly in the presence and number of free-ending veins (FEVs; Fig. 2a–d). The morphology of the cells in the FEVs is diverse, often including one to three tracheids with large lumens, and cell walls thickened in discrete hoops (Strain, 1933). FEVs may account for a significant portion of VLA, and the number of FEVs correlates with the total VLA across Arabidopsis thaliana genotypes and across diverse species (Scoffoni et al., 2011; Dhondt et al., 2012). The FEVs may in fact contribute disproportionately to K_leaf, as pathways for bulk water flow, and represent a substantial evaporative surface area. In most species, FEVs have little or no phloem inside (Van Bel, 1993); in the few species that do, VLA tends to be lower and the FEVs might facilitate sugar export (Adams et al., 2007).

8. Bundle sheath and bundle sheath extensions (BSEs)

The BS surrounding the veins is made up of cells distinct from surrounding mesophyll cells in their elongate morphology, gene expression and chloroplast development (Kinsman & Pyke, 1998). The BS plays numerous roles, including water and nutrient transport via channels and pumps, phloem loading and water storage (Leegood, 2008; Griffiths et al., 2012). In maize, BS development involved gene expression in common with the root endodermis (Slewinski et al., 2012). The BS may be a locus for the sensitivity of K_leaf to decline with dehydration (Shatil-Cohen et al., 2011; Pantin et al., 2013). C₄ species typically have ‘Kranz anatomy’, that is, enlarged BS cells dense with chloroplasts for CO₂ assimilation (Sage, 2004).

Bundle sheath extensions are achlorophyllous cells that in some species extend from the BS to the epidermis, and separate the leaf into chambers. Leaves with relatively dense BSEs are known as ‘heterobaric’ (Fig. 2m,n), and those without are known as ‘homobaric’ (Fig. 2l,o). BSEs have important roles (Notes S6), influencing light absorption (Nikolopoulos et al., 2002; Liakoura et al., 2009); K_leaf and its dynamics with irradiance and leaf water status (Zwieniecki et al., 2007; Scoffoni et al., 2008; Sommerville et al., 2012; stomatal control (Terashima, 1992; Buckley et al., 2011); and biomechanical support (Haberlandt, 1914; Read & Stokes, 2006). A function of BSEs in sugar transport has not been supported by their anatomy or chemical composition; BSEs may have few plasmodesmatal connections with other cells, and their low osmotic concentrations point to a low sugar content (Russin & Evert, 1985b).

9. Accessory transport and support tissues: sclereids, transfusion tracheids and idioblasts

Numerous conifers, cycads and angiosperms have specialized lamina tissue beyond the vein network, including sclereids (Fig. 2g), transfusion tracheids (Fig. 2l,m), fibers or idioblasts; these may be isolated, or form a network throughout the lamina. These tissues may also guide light to deeper mesophyll layers (Karabourniotis, 1998); contribute to higher K_leaf by conducting water, or by providing additional evaporative surface; shorten the hydraulic pathways; and/or contribute to water storage (Brodribb et al., 2010). The transfusion tracheids that radiate from the midvein in conifer needles, or extend from the midrib parallel to the leaf surface in podocarps and in cycads, provide a higher K_leaf than expected solely from VLA (Brodribb et al., 2007). This tissue may also contribute biomechanical support, especially during dehydration, by reducing leaf shrinkage (Cutler, 2005).

10. Leaf size

Leaf size is developmentally related to the venation architecture (see Section 14). Larger leaves tend to have larger petioles and major veins, which contain more numerous and larger xylem and phloem conduits, allowing transport capacity per unit leaf area to be independent of leaf size (Fig. 1a–d; McKown et al., 2010). However, the minor vein diameter and cell sizes tend to be independent of leaf size (Sack et al., 2012). These scaling trends allow K_leaf and gas exchange to be independent of leaf size, and to be separately optimized to the environment (Sack et al., 2012).

Larger leaves may be intrinsically vulnerable to drought or freeze–thaw embolism, given their lower major VLA and their larger xylem conduit diameters. The venation architecture might limit maximum leaf size in certain cases, if low major VLA renders the leaf especially vulnerable to drought or overheating. In single-veined conifers, lamina area is linked with environmental moisture as expected from a direct hydraulic limitation of size (Zwieniecki et al., 2004a; Brodribb & Feild, 2008). This can be remedied with multiple veins, or by the accessory transport and support tissues (Brodribb et al., 2010).

The greater sizes of major veins in larger leaves contributes to their higher LMA, that is, the ‘diminishing returns’ in leaf area per mass invested in larger leaves (Grubb, 1998; Niinemets et al., 2007a; Niklas et al., 2007; Poorter et al., 2009; Sack et al., 2012).

1. Overall framework of vein development during leaf expansion

The algorithm of angiosperm vein development is very general, from A. thaliana to a wide range of diverse eudicotyledons (Sack et al., 2012). This generality indicates genetic obstacles to evolution and redundant developmental pathways (Kang & Dengler, 2004; Kang et al., 2007), consistent with the finding of low number of vein mutants in natural A. thaliana ecotypes and mutagenized populations (Candela et al., 1999).

Vein development begins as the primordium takes shape. Indeed, vascular parenchyma might act as driver cells, providing the impetus for leaf expansion (Van Volkenburgh, 1999). Within the ground meristem, vascular identity gene expression and signal transduction begin the process of vein formation, followed by the differentiation of procambial strands (long cells with few organelles and dense cytoplasm) from central ground meristem cells. The developing vein pattern can be assessed once procambial cells are identifiable (Nelson & Dengler, 1997; Scarpella et al., 2006) or earlier by visualizing the ‘preprocambial’ markers of gene expression (Kang & Dengler, 2004; Scarpella et al., 2004; Sawchuk et al., 2007; Notes S7).

Veins of given orders form in specific ‘windows’ of development time. Leaf development can be separated into two phases, a ‘slow’, limited expansion phase principally as a result of cell proliferation, and a ‘rapid’ and dramatic increase phase resulting mainly from cell expansion, although cell divisions continue. The first- and second-order veins are formed during the slow leaf expansion phase, the third-order veins next, and the minor veins principally during the rapid phase (Sack et al., 2012). In the typical sequence in A. thaliana, the leaf and midvein are initiated by preprocambial signals within the surface of the shoot apical meristem (SAM) by polar auxin transport (Notes S7). Marginal cells express the auxin efflux carrier protein PIN1 asymmetrically, that is, predominantly on one side, polarized toward a ‘convergence point’ on the SAM epidermis, causing it to accumulate auxin (Scarpella et al., 2010). From this convergence point, the midvein preprocambium appears: a basipetally developing strand of cells with PIN1 expressed asymmetrically along their basal or inwardly directed cell surfaces, and this developing strand extends linearly into the SAM, where it signals the placement of the primordium and thus contributes to phyllotaxis (Sawchuk et al., 2007; Bayer et al., 2009; Scarpella et al., 2010). Next, the midvein procambium develops progressively along that strand of cells, extending acropetally from the SAM into the leaf priomordium as it develops (Nelson & Dengler, 1997). The second-order veins form as the marginal meristem grows outward. Preprocambial genes are expressed in the dermal tissue at the leaf margin, and again PIN1 carriers are oriented toward convergence points that accumulate auxin; the preprocambial signal for the second-order veins begins from these convergence points, and spreads in lines of cells from the margin toward the midvein, incrementally recruiting ground meristem cells at their terminus (Scarpella et al., 2006; Sawchuk et al., 2007). The procambium then differentiates from the midrib to the margin progressively along those paths (Kang & Dengler, 2004; Scarpella et al., 2004).

Higher-order veins form in between lower-order veins as the leaf expands. The initiation of procambial strands for each vein order is limited by the need to maintain a critical cell number or distance between new strands and by the duration that the ground meristem can perceive vascular-forming signals (Sack et al., 2012). The preprocambium for each minor vein order appears as cell files throughout the lamina, each extending in two directions from a cell with a bipolar distribution of PIN1 toward an earlier formed vein; the procambium for each vein order differentiates from those cell files (Scarpella et al., 2006, 2010). Leaf expansion proceeds simultaneously, such that the VLA of each order peaks as procambium forms, then declines as leaf expansion pushes veins apart, but in A. thaliana the minor VLA stabilizes as minor veins continue to be initiated in between pre-existing vein strands during ongoing expansion (Kang & Dengler, 2004; Kang et al., 2007). In some species, after all vein strands are laid down, the leaf continues to expand, causing a decline in total VLA (Sack et al., 2012). The FEVs initiate separately in the mesophyll, and grow to attach in the minor vein network. This general pattern of vein formation is superimposed on a basipetal wave of lamina expansion, so that higher orders of venation are formed earlier in the apical portion of the leaf and later in the basal portion (Nelson & Dengler, 1997; Scarpella et al., 2006). Termination of the initiation of procambium from ground meristem is linked to mesophyll cell differentiation (Scarpella et al., 2004; Kang et al., 2007).

Vein diameter expansion depends on vein order. The first- and second-order veins have prolonged diameter growth, whereas third-order veins and minor veins rapidly attain maximum diameter (Nelson & Dengler, 1997; Sack et al., 2012; Taneda & Terashima, 2012). In herbs, the thickening of minor veins is the result of cell proliferation, but in woody species, the first- and second-order veins can thicken by cambial growth (Plymale & Wylie, 1944). In some species, second-order veins may develop branches that form loops later in development; in other species, third-order veins can be thickened and thus transformed into second-order veins, but this accounts for a small portion of the network.

Finally, the phloem and then the xylem mature within the veins (Esau, 1977; Scarpella et al., 2004; Sawchuk et al., 2007). The phloem of the major veins functions in unloading sugars for leaf development, until the minor veins are fully formed, and then phloem unloading ceases, as the minor veins become functional for phloem loading (Turgeon, 2006).

In monocots with striate venation (and in eudicots such as pitcher plants; Franck, 1976), a similar process unfolds. The large longitudinal veins develop within the primordium, and as the leaf expands, these veins are spaced apart and the small longitudinal veins form in between, followed by the transverse veins (Denne, 1960). In grasses, the small longitudinal veins are connected apically and basally with adjacent veins within the leaf blade, and do not extend into the sheath region. Dicotyledons with exceptional venation also exhibit departures from the typical developmental sequence. Thus, in Calophyllum spp., the dense, parallel second-order veins branching from the midrib develop during the rapid expansion phase, as is the case for minor veins in the typical dicot leaf (Ramji, 1967).

The major and minor veins differ not only in the timing of development, but also in their gene expression during development, procambial anatomy, and xylem and phloem formation, just as they differ in many aspects of their function and evolution (see Table S5 in Sack et al., 2012).

The BS tissue differentiates early with the developing vascular strand and cells divide and elongate in concert with vascular cells (Kinsman & Pyke, 1998). In species with multilayered BSs (e.g. grasses, which have an inner ‘mesome sheath’), the inner layer can arise from the procambium, whereas the outer layers may arise from the ground meristem, which produces the mesophyll tissue (Bosabalidis et al., 1994; Trivett & Evert, 1998). Accessory xylem tissues tend to form after the vein system is fully differentiated and the mesophyll has matured. Sclereids and mesophyll fibers can form independently from initials that appear late in mesophyll tissue differentiation (Arzee, 1953; Tomlinson & Fisher, 2005).

2. Coordination of vein traits with other cell and leaf traits in development and evolution

The developmental algorithm for vein formation during leaf expansion provides a basis for trait correlations within and across species. First, there are trait correlations within given leaves. Thus, in many species, the second-order veins are spaced further apart in the center of leaves, which expands most; their spacing can approximate the appearance of wavefronts, with their separation corresponding to development time (Pietak, 2009; Rolland-Lagan et al., 2009). Further, because cell expansion after veins are patterned reduces minor VLA, areas of the leaf with smaller cells have higher VLA, for example, the tips and margins (Zwieniecki et al., 2004b; Nardini et al., 2008).

Across species, vein traits scale characteristically with mature leaf size as predicted by the developmental algorithm. Across a global dataset, major VLA declined with final leaf size, and major vein diameter increased with leaf size; this is because major vein formation occurs in the slow phase of leaf expansion, and these veins are subsequently pushed apart, although they can be thickened during most of their development. By contrast, minor VLA and diameter were independent of leaf size, as these can be initiated progressively during rapid leaf expansion, and rapidly achieve final diameter (Sack et al., 2012; Notes S8). The scaling trends would be reinforced by additional adaptation to the environment. For example, when a larger leaf is selected for a given ecological advantage, this would reinforce the global scaling trends, because that leaf would automatically have a lower major VLA, according to the developmental mechanism. Further, this larger leaf would require a larger midrib for biomechanical and hydraulic support, and that would arise from the linked development of midrib diameter and leaf size. Selection for higher or lower minor VLA, independently of the final leaf size, would reinforce the independence of VLA from leaf size (Sack et al., 2012). Vein chemistry also differs across the venation network, as expected from the developmental algorithm. The first- and second-order veins have less negative carbon isotope ratios than lamina, indicating preferential allocation of stored ‘sink’ phloem sugars rather than ‘source’ sugars in the growing leaf, as the major veins develop while the leaf is a sink (Badeck et al., 2009).

The functional coordination of vein traits and stomatal traits (6) also has a genetic and developmental basis. Studies of the fossil record and of modern adaptive radiations have shown correlations of vein traits and stomatal traits in evolution (Upchurch, 1984; Dunbar-Co et al., 2009; Zhang et al., 2012). Mutant studies also suggested that vein and stomatal patterning overlap in genetic basis (Jover-Gil et al., 2012). Stomatal density is determined during the same development time as the venation (Pantin et al., 2012) and in the mature leaf depends on epidermal cell numbers and sizes. Arabidopsis thaliana genotypes with larger cells at leaf maturity tend to have both lower VLA and lower stomatal density (Manuel Perez-Perez et al., 2011). Similarly, in some species the sun-shade plasticity of vein traits is linked with shifts in cell and leaf size. Shade leaves of many species have greater cell and leaf expansion after the vein strands are fully formed, and thus develop a lower VLA and stomatal density than sun leaves (Brodribb & Jordan, 2011; Murphy et al., 2012). However, this developmental coordination of vein and epidermal traits can be decoupled by shifting the amount and/or timing of vein development and stomatal initiation relative to cell and leaf expansion. Thus, in other species, shade leaves may develop lower, or similar, VLA, with larger, similar-sized, or even smaller leaves (Sack et al., 2012).

Vein development can also set the pattern for other specialized tissues. In Lithops, the vein areoles guide the formation of unique epidermal ‘windows’ for light entry (Korn, 2011).

1. Plasticity of vein traits within canopies and across environments

Vein traits, like other leaf traits, including leaf size and stomatal density, are determined by signals before and during leaf development (Zwieniecki et al., 2004b; Pantin et al., 2012). Many studies have demonstrated the plasticity of vein traits within canopies and for plants of given species across environments. Theoretically, leaves acclimated to higher irradiance, temperatures and nutrient supplies would have vein traits associated with higher K_leaf and rates of gas exchange (Table 1). However, leaves acclimated to lower water supply would also often have vein traits associated with higher K_leaf and rates of gas exchange, and also with greater drought tolerance. Such leaves can better ‘gear up’ to higher function when water is available, and to ‘gear down’ with stomatal closure and/or dormancy during periods of extended water shortage (Maximov, 1931; Grubb, 1998; Scoffoni et al., 2011). This theory was supported by our synthesis of studies of vein trait plasticity (Table S1).

2. Variation in vein traits with growth form, habitat and biomes

Species adaptation across habitats drives strong shifts in leaf vein traits. Adaptation followed the same patterns described for plasticity across environments in a synthesis of studies of species variation in vein traits across habitats, for species within and across lineages (Table S1).

As a first test for worldwide trends in vein adaptation to climate, we compiled global data on VLA for 796 species (Table S3). This analysis resulted in critical novel trends (Figs 4, 5), consistent with those reported so far for smaller species sets and individual lineages (Table S1). Within given biomes and growth forms, deciduous species had higher mean VLA than evergreens. Further, within biomes, mean VLA increased from herbs to shrubs to trees, paralleling a trend previously shown for stomatal density (Beaulieu et al., 2008). Within biomes and growth forms, sun-adapted species had higher VLA than shade-adapted species. Across biomes, there was a strong correlation of mean VLA with aridity (i.e. with the Priestly–Taylor α, an index of moisture supply relative to demand; r_p = −0.93, P = 0.003; inset in Fig. 4). These trends for VLA were also observed when we plotted individual species means against climate variables (Fig. 5); VLA correlated positively with mean annual temperature (MAT) and negatively with mean annual precipitation (MAP) (|r_p| = 0.17, P = 0.03). The VLA also decreased with species’ shade tolerance index. The global trend of increasing mean VLA with aridity matches the smaller-scale trends reported within smaller species sets and individual lineages (Table S1), and supports the idea that, while species vary strongly in their mechanisms of adaptation to aridity, a very common mechanism is to accomplish more photosynthesis during pulses of high water availability (Maximov, 1931; Grubb, 1998; Scoffoni et al., 2011). This trend does not imply that forest species cannot have high VLA. Indeed, pioneer species of wet tropical forests frequently have high VLA (Sack & Frole, 2006; Feild et al., 2011). However, many tropical wet forest canopies are dominated by species that establish as understory shade-tolerators, and their moderate vein densities persist even as they reach the canopy (Sack & Frole, 2006), and would lower the biome mean. This global analysis, based on the substantial available data, provides a baseline against which forests, growth forms, and lineages should be compared.

We also identified shifts in VLA associated with specific functional types, consistent with expectations (Fig. 5). Aquatic herbs had very low VLA, particularly submerged aquatics. Crop herbs and alpine herbs had similar VLA to temperate herbs. Succulent-leafed species had very low VLA, consistent with their low g_s, and, in many cases, CAM photosynthesis. As previously observed, C₄ grasses tended to have higher VLA than C₃ grasses.

Major VLA also showed adaptation to climate and habitat, in part corresponding to its linkage with leaf size (Fig. 5). Leaf size but not major VLA was positively related to MAT. Leaf size also correlated with MAP, as expected from previous meta-analyses, while major VLA declined (r_p = −0.35; P = 0.02). Thus, species of drier habitats tended to have smaller leaves, with greater major VLA, conferring redundancy and drought tolerance (7). Leaves of shade-tolerant species tended to have larger leaves with lower VLA (Fig. 5).

1. Evolution of vein traits in major plant lineages and paleohistory

The leaf venation appeared in the earliest true leaves. The earliest land-plant lineages lacked vasculature, although some may have had analogous structures. Certain mosses evolved laminate structures that deliver some water through hydroids in their costae. The multiple evolution of leaves is analogous to that of wood, which evolved five or more times by the mid-Carboniferous (320 million yr ago (Mya)). In the early Devonian, c. 400 Mya, leaf-like organs evolved independently in two lineages. In lycophytes, microphylls (containing a single vascular bundle) evolved through enations, small outgrowths of tissue later becoming vascularized, or through the sterilization of sporangia (Crane & Kenrick, 1997; Floyd & Bowman, 2006). In euphyllophytes, megaphylls probably originated multiple times as lateral branches (Floyd & Bowman, 2006; Tomescu, 2009). This phenomenon has been hypothesized to involve infilling (‘webbing’) between side branches on the same plane, according to Zimmermann's 1930 telome theory (Wilson, 1953). Webbing may arise as a continuation of meristematic activity as a marginal meristem, rather than splitting of meristems during dichotomy in the ancestral branched system (Doyle, 1998; Beerling & Fleming, 2007).

Vein reticulation and hierarchy, simpler than that for angiosperms, emerged in earlier lineages via different evolutionary and developmental pathways (Table S2; Alvin & Chaloner, 1970; Trivett & Pigg, 1996). We provide a synthesis of leaf venation in the major plant groups, highlighting vein complexity, reticulation and presence of FEVs and BSEs (Fig. 6; Table S2).

2. Vein traits as key innovations in angiosperms

Angiosperms evolved numerous distinctive vein traits contributing to their greater performance than earlier-evolved lineages. Early angiosperms tended to have low-order veins with less organization, correlating with stomata varying strongly in the numbers and arrangement (Upchurch, 1984). Later evolution of higher VLA (> 5 mm mm⁻²; Boyce et al., 2009; Brodribb & Feild, 2010; Feild et al., 2011) corresponded with increasing numbers of vein orders (Brodribb & Feild, 2010) and higher stomatal density (Franks & Beerling, 2009). Angiosperms required the evolution of the plate meristem to produce higher vein orders (Doyle & Hickey, 1976; Boyce, 2008; Sack et al., 2012). A high VLA would also require the evolution of a vein hierarchy and vessels, because a high-VLA system built of large veins comprising tracheids would be prohibitively expensive (Beerling & Franks, 2010; McKown et al., 2010). Further, the evolution of simple perforation plates from scalariform plates may have been essential (Feild & Wilson, 2012). Sieve tubes may also have been a necessary advance. In Amborella trichopoda, sister to all other angiosperms, the minor veins showed anatomical and physiological specialization for both apoplastic and symplastic loading via polymer trapping, indicating versatile efficiency (Turgeon & Medville, 2011).

These innovations made possible large leaves with large major veins for mechanical support and a high VLA that enables transpirational cooling and high photosynthetic rates (Niklas, 1999; Osborne et al., 2004; Boyce, 2008; Brodribb et al., 2010; Walls, 2011; Sack et al., 2012). Indeed, the angiosperm system allows for a dramatic range of leaf sizes with a wide spectrum of VLA, vein conductivities, and photosynthetic rates (Wright et al., 2004), and enables dominance in a far greater range of habitats than other plant lineages (Boyce et al., 2009; Brodribb et al., 2010; McKown et al., 2010, 2010; Peppe et al., 2011).

This advanced venation, in addition to a more sophisticated stomatal control system and the floral organ, set the stage for the rise of angiosperms to global dominance.

3. Links of vein evolution with atmosphere and climate and influence on the biogeography of key vegetation types

The evolution of venation, like that of wood and leaves, may have been critically coupled to changes in atmospheric CO₂ (Sperry, 2003; Brodribb & Feild, 2010; de Boer et al., 2012). Indeed, the evolution of megaphylls is linked to the drop in atmospheric CO₂ in the late Devonian (Beerling et al., 2001; Fig. 6). Megaphylls evolved higher stomatal densities, enabling the fixation of sufficient CO₂ to survive, and transpiration rates adequate to cool their planate laminae (Beerling et al., 2001; Shougang et al., 2003; Osborne et al., 2004). Similarly, the evolution of high VLA in angiosperms may have been precipitated by declining CO_2, enabling greater stomatal apertures and thus maintaining or increasing photosynthetic performance (Beerling & Franks, 2010; de Boer et al., 2012), providing a strong advantage over competing plant groups (Brodribb & Feild, 2010; Feild et al., 2011; Fig. 6). There is some debate about the ability of VLA to drive differences among groups in photosynthetic rates as atmospheric CO₂ changed, with predictions depending on the traits and assumptions used to estimate A_area from VLA (de Boer et al., 2012; Boyce & Zwieniecki, 2012).

The leaf venation may have impacts on the distribution of vegetation systems. The evolution in angiosperms of higher VLA and transpiration rates may have driven the establishment of angiosperm-dominated rainforests, and influenced the global water budget (Boyce et al., 2009, 2010; Boyce & Lee, 2010). Vein traits also likely played a role in the evolution of grasslands. Higher VLA in grasses probably evolved in response to declining CO₂ in the Miocene and with the colonization of drier, more exposed habitats (Edwards et al., 2010). High VLA probably potentiated the repeated evolution of C₄ and enabled the emergence of this key biome (Sage, 2004; Osborne & Sack, 2012).