Leaf shape and venation pattern alter the support investments within leaf lamina in temperate species: a neglected source of leaf physiological differentiation?

Authors

  • Ü. NIINEMETS,

    Corresponding author
    1. Department of Plant Physiology, University of Tartu, Riia 23, Tartu 51010, Estonia, and
    2. Institute of Agricultural and Environmental Sciences, Estonian University of Life Sciences, Kreutzwaldi 64, Tartu 51014, Estonia
    Search for more papers by this author
  • A. PORTSMUTH,

    1. Department of Plant Physiology, University of Tartu, Riia 23, Tartu 51010, Estonia, and
    Search for more papers by this author
  • M. TOBIAS

    1. Department of Plant Physiology, University of Tartu, Riia 23, Tartu 51010, Estonia, and
    Search for more papers by this author

†Author to whom correspondence should be addressed. E-mail: ylon@ut.ee

Summary

  • 1A trade-off between the investments in functional and support structures is a major determinant of leaf physiological activities. A variety of leaf shapes and venation densities occur in coexisting vegetation, but the costs and benefits of various leaf shapes and venation architectures are poorly understood. As the lever arms (location of leaf mass centre) become effectively longer as the maximum of lamina mass distribution shifts farther from the lamina base, we hypothesized that the fraction of lamina biomass in the mid-rib (FMR) is larger in leaves in which the centroid of lamina mass is located at, or greater than, half-leaf length (‘elliptic’ leaves) compared with leaves having the centroid of lamina mass located closer to the leaf base (‘ovate’ leaves). We further hypothesized that minor vein density (ρV) is larger in leaves with lower FMR, compensating for lower investments in central support. Finally, we predicted that ρV is lower in parallel/palmate-veined than in pinnate-veined leaves, due to a more uniform distribution of large veins in parallel/palmate-veined leaves.
  • 2FMR and ρV were studied in 44 herbs and woody seedlings with an overall variation in lamina fresh mass (MFL) of more than five orders of magnitude, and a sixfold variation in leaf longevity. Species were separated between pinnate-veined elliptic and ovate leaves, and parallel- or palmate-veined elliptic and ovate leaves.
  • 3Contrary to the hypothesis, support investment in the mid-rib was similar among leaf shapes, and scaled positively with leaf size and negatively with leaf longevity. However, FMR and ρV were negatively associated. Fractional biomass investment in the mid-rib scaled with lamina size (fresh and dry mass and area), but at a common lamina size FMR was larger in pinnate-veined elliptic than in parallel/palmate-veined elliptic leaves. In addition, ρV was larger in pinnate than in parallel/palmate-veined leaves, and the differences in lamina carbon content further suggested an overall greater investment of lamina biomass in the minor veins of pinnate-veined leaves.
  • 4These data demonstrate that the effect of leaf shape on biomass investments in central support is less significant than predicted by biomechanical models, partly because of the trade-off between the biomass investments in central support and minor veins, which compensate for differences in lamina shape. These data collectively indicate that leaf size, longevity, shape and venation pattern can importantly modify the distribution of foliage biomass between support and functional tissues, and thus can alter foliage physiological activity and leaf functioning in environments with different resource availability.

Introduction

Plant species with widely varying leaf shape, size and venation architecture co-occur in vegetation. While the differences in foliar physiognomy have attracted numerous researchers, the significance of different leaf shapes and sizes for species niche differentiation is still not entirely understood, partly because of a lack of information on the costs and benefits of specific leaf forms. Variations in leaf size are typically explained by optimization of temperature responses of leaf gas exchange in different environments (for review see Givnish 1978, 1987). Apart from gas exchange, large-leaved species appear to require less stem xylem to support the same leaf area (Pickup et al. 2005), suggesting an advantage for larger-leaved species in humid, non-stressful environments where the influence of temperature on carbon gain is minor. However, mechanical theory predicts that, for leaves of common thickness, the fraction of within-leaf biomass invested in support should be greater in larger leaves due to effectively longer lever arms (Givnish 1978; Gere & Timoshenko 1997; Niinemets & Fleck 2002), possibly offsetting the advantages of the low stem xylem requirement of larger leaves (Pickup et al. 2005). Within-leaf support investments have been investigated in a few species and generally demonstrate a scaling of biomass investment in the mid-rib with leaf size, with several exceptions and large interspecific variability (Givnish 1986; Niinemets & Sack 2006).

One factor potentially altering the relationships between the fraction of leaf biomass in support and leaf size is leaf shape. Leaf shape is strongly linked with vascular patterns and biomass investments in vasculature (Dengler & Kang 2001; Kessler & Sinha 2004). In particular, the way leaf lamina mass is distributed from leaf base to apex determines the overall bending moment and the support investment required by leaves of common area and mass. The support requirements are apparently less for cordate and ovate leaves, in which most leaf mass is located close to the leaf base, than for elliptic or obovate leaves, in which the bulk of leaf mass is located farther from the leaf base. Optimization of water supply and biomechanical support for a given biomass investment in the mid-rib suggests that the optimal leaf form is wedge-shaped (Givnish 1978). Although these leaves are very efficient in terms of support, most angiosperm leaves are not wedge-shaped, possibly partly because wedge-shaped leaves cannot be packed efficiently on the stem (Givnish 1978). In addition, allocating most of the leaf mass and the light-intercepting area further from the point of axial support reduces self-shading and increases leaf light interception, potentially enhancing leaf carbon gain (Pearcy & Yang 1998; Pearcy et al. 2005).

There are fundamental relationships between leaf physiological and structural characteristics that are apparently valid for leaves of widely different physiognomy (Wright et al. 2004). Negative relationships between leaf photosynthetic capacity per dry mass and leaf dry mass per unit area arise as leaves with more robust structures have a greater fraction of support tissues within the leaves, less efficient internal diffusion, and lower fractions of nitrogen invested in photosynthetic enzymes (Reich et al. 1995; Poorter & Evans 1998; Niinemets 1999b; Wright et al. 2004; Terashima et al. 2005; Niinemets & Sack 2006; Niinemets et al. 2006a). However, leaves of similar thickness and lamina structure, but of different size, may have widely different investments of foliar biomass in the mid-rib. Since the mid-rib has lower photosynthetic activity compared with the rest of the lamina, the contrasting within-leaf support investments among leaves of varying shape and size can significantly modify the scaling relationships between leaf structure and physiology at any specific point along the fundamental spectrum of leaf functioning. This study aimed to test the hypothesis that within-leaf support costs are greater for leaves with larger area and mass; and that for leaves with equivalent area and mass, the support costs are larger for elliptic and obovate leaves than for ovate and cordate leaves (Fig. 1).

Figure 1.

Outline of working hypotheses of how leaf shape and venation affect biomass investment in support. For leaves of the same length (L) the location of mass centre (centroid, c) varies in dependence on the distribution of leaf mass along the leaf length. We hypothesize that the fraction of biomass in the mid-rib required for leaf self-support is larger for leaves with biomass located at a distance x(c) ≥ L/2 (hypothesis 1). As large veins are distributed more uniformly in the leaf lamina for palmate- and parallel-veined than for pinnate-veined leaves, we further assume that the density of minor veins (ρV, length per unit area) is larger for pinnate-veined leaves (hypothesis 2). The distance of mass-centre (centroid) location, x(c), relative to leaf length is 0·34 for the depicted cordate, 0·4 for the ovate, 0·52 for the elliptic and 0·61 for the obovate leaf.

Apart from the mid-rib, a large fraction of leaf biomass can be invested in minor veins, which also play an important role in mechanical support and water conduction. In many species, a major part of leaf hydraulic resistance is in the minor veins (Sack et al. 2004, 2005). As the density of major veins is larger for palmate- and parallel-veined than for pinnate-veined leaves, and veins with a larger cross-sectional lumen area are hydraulically more efficient (Sack & Holbrook 2006), within-leaf costs for production of minor and major veins may be inversely associated. In addition, palmate- and parallel-veined leaves may also be mechanically more efficient for a given biomass investment in major veins (Roth-Nebelsick et al. 2001), especially palmate- and parallel-veined leaves with cordate and ovate shapes that have a major part of the leaf biomass located close to the base. We tested the hypotheses that pinnate-veined leaves have a higher density of minor veins than parallel- or palmate-veined leaves; and that there is an overall negative relationship between minor vein density and biomass investment in the mid-rib (Fig. 1).

In many plant species, turgor (the pressure exerted by cellular water on cell walls) contributes significantly to leaf support, in particular in species with thin cell walls and high leaf water content (Niklas 1986; Niklas 1989), which also tend to have a shorter life span and greater physiological activity (Reich et al. 1999; Wright et al. 2004). As species with higher leaf water content are expected to rely more strongly on turgor for mechanical support, we hypothesized further that among leaves of common mass, area and shape, dry matter investment in the mid-rib is less for short-lived leaves with relatively higher leaf water content.

Materials and methods

study sites and plant sampling

The study was conducted at the end of June 2000 in Estonia. Within-leaf and whole-plant biomass distribution was studied in 26 herb species and in seedlings of six woody species (Table 1) in a broad-leaved deciduous forest in Ülenurme (58°18′ N, 26°42′ E, elevation 60 m). The soil is a sandy-loam glossic gleysol (FAO, ISRIC & ISSS 1998; pseudopodzolic soil, cf. Kõlli & Ellermäe 2003 for a comparison between FAO and Estonian soil classifications). The forest overstorey is dominated by 16–19-m-tall trees of Populus tremula L. and Fraxinus excelsior L., while the shrub layer is dominated by Corylus avellana L. and the herb layer by Calamagrostis arundinacea (L.) Roth, Fragaria vesca L., Hepatica nobilis Gras., Melampyrum nemorosum L. and Oxalis acetosella L. A detailed description of the study site is provided by Niinemets & Kull (1998).

Table 1.  Leaf habit, life span and average (± SE) lamina fresh mass (MFL)*, dry mass per unit area (MA), dry to fresh mass ratio (DF) and nitrogen content per dry mass (NL) in 33 species sampled from the broad-leaved mixed deciduous forest (F) and in 11 species sampled from the botanical garden (G)
SpeciesFamilySiteLeaf/leaflet shapeVenation typeLeaf life span (months)MFL (g)*MA (g m−2)DF (g g−1)NL (%)
  • Corylus avellana, Lonicera xylosteum, Ribes alpinum, Ribes nigrum and Rubus idaeus are shrubs; Fraxinus excelsior, Padus avium, Populus tremula and Sorbus aucuparia are trees; all other species are herbs.

  • *

    Lamina FM includes mid-rib and the rest of the lamina. For compound-leaved species (Aegopodium podagraria, Angelica sylvestris, Anthriscus sylvestris, Fragaria vesca, Fraxinus excelsior, Geranium palustre, Oxalis acetosella, Paris quadrifolia, Rubus idaeus, Rubus saxatilis, Sorbus aucuparia), average lamina FM of leaflets is provided.

Aegopodium podagraria L.ApiaceaeFOvatePinnate3·50·159 ± 0·01229·2 ± 1·70·293 ± 0·0142·02 ± 0·19
Angelica sylvestris L.ApiaceaeGOvatePinnate3·50·15 ± 0·0819·1 ± 1·90·255 ± 0·0292·61 ± 0·21
Anthriscus sylvestris (L.) Hoffm.ApiaceaeFOvatePinnate3·50·0129 ± 0·00116·2 ± 1·90·256 ± 0·0292·72 ± 0·34
Arctium tomentosum Mill.CompositaeGCordatePinnate3·56·0 ± 2·341 ± 70·175 ± 0·0452·5 ± 0·7
Armoracia rusticana (Lam.) Gaer.CruciferaeGEllipticPinnate246 ± 2024·8 ± 3·50·18 ± 0·053·43 ± 0·48
Asarum europaeum L.AristolochiaceaeFReniformParallel130·413 ± 0·04744·4 ± 2·90·231 ± 0·0142·24 ± 0·17
Cirsium oleraceum (L.) Scop.CompositaeFOvatePinnate3·510·0 ± 3·522·5 ± 2·30·101 ± 0·0293·04 ± 0·35
Convallaria majalis L.LiliaceaeFEllipticParallel3·50·42 ± 0·0844·2 ± 3·70·31 ± 0·052·15 ± 0·42
Corylus avellana L.CorylaceaeFOval/obovatePinnate70·23 ± 0·1029 ± 50·36 ± 0·062·14 ± 0·25
Crepis paludosa (L.) MoenchCompositaeFObovatePinnate3·50·37 ± 0·3513·8 ± 1·30·165 ± 0·0141·97 ± 0·22
Cucurbita pepo L.CucurbitaceaeGOrbicularPalmate257 ± 1627·7 ± 4·50·083 ± 0·0126·6 ± 0·5
Echinops sphaerocephalus L.CompositaeGObovatePinnate3·57·8 ± 1·247 ± 70·189 ± 0·0373·25 ± 0·36
Epilobium montanum L.OnagraceaeFOvatePinnate3·50·17 ± 0·0722·3 ± 4·60·112 ± 0·0324·4 ± 0·6
Fragaria vesca L.RosaceaeFOvalPinnate3·50·056 ± 0·00932 ± 80·414 ± 0·0291·92 ± 0·04
Fraxinus excelsior L.OleaceaeFEllipticPinnate70·029 ± 0·00918·0 ± 2·60·266 ± 0·0181·60 ± 0·27
Galeobdolon luteum Huds.LabiataeFOvatePinnate3·50·101 ± 0·04735 ± 100·262 ± 0·0362·45 ± 0·12
Geranium palustre L.GeraniaceaeFPalmatePalmate3·50·63 ± 0·0623 ± 60·226 ± 0·0242·16 ± 0·28
Helianthus annuus L.CompositaeGCordatePalmate238 ± 1035·3 ± 2·70·111 ± 0·0294·7 ± 1·1
Hepatica nobilis Gars.RanunculaceaeFOrbicularPalmate130·370 ± 0·02946·8 ± 2·20·320 ± 0·0101·91 ± 0·09
Heracleum sosnowskyi MandenApiaceaeGOrbicular/ovatePinnate230 ± 1231 ± 50·167 ± 0·0452·4 ± 0·5
Ligularia wilsoniana (Hems.) Green.CompositaeGCordatePinnate3·526 ± 1436·4 ± 2·90·19 ± 0·051·53 ± 0·19
Lonicera xylosteum L.CaprifoliaceaeFOvalPinnate70·12 ± 0·0547 ± 170·329 ± 0·0171·97 ± 0·45
Maianthemum bifolium (L.) Schm.LiliaceaeFCordateParallel3·50·18 ± 0·0535·8 ± 2·90·251 ± 0·0351·33 ± 0·04
Melampyrum nemorosum L.ScrophulariaceaeFOvatePinnate3·50·0565 ± 0·04529·3 ± 4·50·237 ± 0·0152·93 ± 0·45
Oxalis acetosella L.OxalidaceaeFObcordateParallel60·033 ± 0·02612·9 ± 3·00·103 ± 0·0472·39 ± 0·48
Padus avium Mill.RosaceaeFEllipticPinnate70·20 ± 0·0641 ± 110·41 ± 0·091·98 ± 0·35
Paris quadrifolia L.LiliaceaeFObovateParallel3·50·18 ± 0·0516·3 ± 3·60·158 ± 0·0242·04 ± 0·18
Polygonatum multiflorum (L.) All.LiliaceaeFOvalParallel3·50·46 ± 0·0928·4 ± 2·80·21 ± 0·062·42 ± 0·15
Polygonum cuspidatum Sieb. et Zucc.PolygonaceaeGOvatePinnate74·0 ± 1·356 ± 90·32 ± 0·071·70 ± 0·50
Populus tremula L.SalicaceaeFOrbicularPinnate70·19 ± 0·0626 ± 50·254 ± 0·0393·5 ± 0·8
Prunella vulgaris L.LabiataeFOvatePinnate3·50·063 ± 0·01516·7 ± 1·90·181 ± 0·0271·82 ± 0·43
Ranunculus cassubicus L.RanunculaceaeFReniformPalmate3·50·87 ± 0·2144·6 ± 1·60·281 ± 0·0492·37 ± 0·41
Rheum rhabarbarum L.PolygonaceaeGCordatePalmate2208 ± 9234·3 ± 1·80·066 ± 0·0023·74 ± 0·23
Ribes alpinum L.GrossulariaceaeFOvatePalmate70·136 ± 0·01942 ± 110·276 ± 0·0212·0 ± 0·7
Ribes nigrum L.GrossulariaceaeFOvatePalmate70·253 ± 0·04534·5 ± 3·50·360 ± 0·0111·32 ± 0·43
Rubus idaeus L.RosaceaeFOvatePinnate70·043 ± 0·01837 ± 70·28 ± 0·072·98 ± 0·49
Rubus saxatilis L.RosaceaeFOvatePinnate 3·50·20 ± 0·0622·9 ± 3·90·261 ± 0·0431·49 ± 0·34
Rumex crispus L.PolygonaceaeGOvatePinnate3·58·3 ± 4·245·6 ± 2·60·280 ± 0·0342·20 ± 0·15
Sorbus aucuparia L.RosaceaeFEllipticPinnate70·022 ± 0·01830 ± 100·416 ± 0·0322·11 ± 0·33
Stellaria holostea L.CaryophyllaceaeFOvate/lanceolatePinnate3·50·048 ± 0·02733·0 ± 4·20·241 ± 0·0121·97 ± 0·42
Taraxacum officinale Web. ex Wigg.CompositaeFObovatePinnate20·72 ± 0·3720·2 ± 3·40·088 ± 0·0142·44 ± 0·33
Tussilago farfara L.CompositaeFCordatePalmate3·52·2 ± 2·033·2 ± 4·10·180 ± 0·0433·22 ± 0·43
Verbascum nigrum L.ScrophulariaceaeFOvatePinnate3·55·2 ± 2·430·9 ± 4·90·115 ± 0·0383·6 ± 0·5
Veronica chamaedrys L.ScrophulariaceaeFOvatePinnate 3·50·034 ± 0·01019·3 ± 3·20·207 ± 0·0231·98 ± 0·09

Above-ground plant material was collected for analysis from at least two individuals of every species (average ± SE = 3·0 ± 0·7). Plants were sampled from medium-sized gaps, and average (± SE) integrated daily quantum flux density for May and June, determined from hemispherical photographs (cf. Niinemets et al. 2004 for details of light estimation), was 8·0 ± 0·8 mol m−2 day−1 corresponding to ≈25% canopy openness. In general, in herb species, light-driven foliage structural and morphological plasticity saturates at a relative irradiance of 20–30% of full light (Winn & Evans 1991; Olff 1992; Niinemets 2004).

The maximum average leaf fresh mass (FM) was ≈20 g in the forest species. To achieve a larger range in leaf FM and increase the generality of our results, within-leaf biomass partitioning was studied in 11 herb species with average leaf FM varying between ≈4 and 320 g (Table 1) at the Botanical Garden of the University of Tartu (58°22′ N, 26°43′ E, elevation 20 m). This site supported an open, deciduous, woody plantation on a sandy-loam mollic cambisol (brown soil), and the average (±SE) integrated daily quantum flux density was 18·1 ± 2·2 mol m−2 day−1, corresponding to ≈50% of full light.

In all cases, the sampled material was enclosed in plastic bags containing wet filter paper and was immediately transported to the laboratory for further analyses.

species studied and species grouping

Species were selected for the study to achieve a representative range of leaves with different size, shape and venation. Of the 44 species sampled (Table 1), 33 were simple-leaved and 11 were compound-leaved (Aegopodium podagraria L., Angelica sylvestris L., Anthriscus sylvestris (L.) Hoffm., F. vesca, F. excelsior, Geranium palustre L., O. acetosella, Paris quadrifolia L., Rubus idaeus L., Rubus saxatilis L., Sorbus aucuparia L.). For the entire data set, variation in average leaf FM was ≈15 000-fold (Table 1); for total within-leaf biomass fraction in support, the variation was approximately eightfold.

Most of the compound-leaved species sampled were simple pinnate, but A. sylvestris was bipinnate; F. vesca and O. acetosella were trifoliate; and G. palustre was palmately compound. In the analyses of leaf shape and venation, leaflets of compound-leaved species were considered as functional analogues of leaves in simple-leaved species (Givnish 1984). The species were divided into two groups (ovate vs elliptic) depending on the location of the centroid (mass centre) of the lamina from the point of attachment (Fig. 1; Table 1). In ‘ovate’-leaved species, the bulk of the leaf mass is located close to the leaf attachment i.e., the lamina mass centre is at a distance a < L/2 from the leaf attachment to the petiole (L is total leaf length). In ‘elliptic’-leaved species, the lamina centroid is located at a distance a ≥ L/2 from the leaf attachment to the petiole (see Table 2 for explanation of variables). Cordate, ovate or palmate shapes were considered ‘ovate’; elliptic, obcordate, oblong, obovate, orbicular, oval or reniform shapes were considered ‘elliptic’ (Table 1).

Table 2.  Variables used in the text
VariableDefinition
ɛ (–)Lamina elongation (ratio of lamina major to minor semi-axes)
γ (–)Lamina (perimeter)2/area
Λ (months)Leaf life span
ρV (mm mm−2)Vein density
a (m)Distance from leaf attachment
CL (%)Lamina (without mid-rib) carbon content per dry mass
DF (g g−1)Lamina (without mid-rib) dry to fresh mass ratio
FMR (g g−1)Mass fraction of mid-rib in lamina, MMR/(ML + MMR)
L (m)Leaf length
MA (g m−2)Lamina (without mid-rib) dry mass per unit area
MAW (g m−2)Total lamina (with mid-rib) dry mass per unit area
MDL (g)Total lamina (with mid-rib) dry mass
MFL (g)Total lamina (with mid-rib) fresh mass
ML (g)Lamina (without mid-rib) dry mass
MMR (g)Mid-rib dry mass
NL (%)Lamina (without mid-rib) nitrogen content per dry mass
NMR (%)Mid-rib N content per dry mass
x(c) (m)Location of lamina mass centre from petiole attachment

Species were grouped according to the venation pattern as pinnate-veined (higher-order veins strongly associated with central vein) and parallel/palmate-veined (higher-order veins less strongly linked to the mid-rib). Leaf venation was pinnate for 28 species, while six species had a parallel venation and 10 species a palmate venation pattern (Table 1). In all comparisons we pooled parallel- and palmate-veined species. Although the parallel and palmate venations have their specific architectural features, the current comparison is justified from a biomechanical perspective as the major veins are distributed more uniformly across the lamina in both parallel- and palmate-veined leaves than in pinnate-veined leaves (Sack et al. 2003a).

The 44 species studied belonged to 20 different families (Table 1); each shape/venation group consisted of species belonging at least to five different families, and all groups contained species belonging to the same families. Thus the species distribution among groups was phylogenetically heterogeneous and the group differences demonstrated in our study are not driven by phylogenetic differences among the groups (sensu Ackerly & Reich 1999).

Seven of the species sampled have been introduced to Estonia from southern Europe (Echinops sphaerocephalus L., Heracleum sosnowskyi Manden, Rheum rhabarbarum L.); temperate East Asia [Japan, China: Ligularia wilsoniana (Hems.) Green, Polygonum cuspidatum Sieb. ex Zucc.]; North America (Helianthus annuus L.) and Central America (Cucurbita pepo L.). Except for C. pepo and H. annuus, all foreign species disperse naturally in Estonia, and may be invasive in suitable sites.

leaf life span

Our data set included deciduous herbs and woody species as well as evergreen herbs (Asarum europaeum L., H. nobilis, O. acetosella; Table 1). Leaf life span (Λ) was estimated using the GLOPNET database, currently the most comprehensive source of leaf life span (Wright et al. 2004), and our own knowledge of species biology. Literature data for deciduous species were adjusted to the length of growing season at our latitudes. Thus woody deciduous species were assigned Λ= 7 months, and most perennial herbs Λ = 3·5 months (7.5–8·5  months for woody deciduous and 4–4·5 months for deciduous perennial herbs in the original GLOPNET database, Wright et al. 2004). Short-lived cultural plants [Armoracia rusticana (Lam.) Gaer., C. pepo, H. annuus, R. rhabarbarum] and invasive herbs (H. sosnowskyi, Taraxacum officinale Web. ex Wigg.) with rapid leaf turnover were assigned a life span of 2 months, while the perennial shrub-like herb P. cuspidatum, which produces only one leaf flush in our climate, was assigned Λ= 7 months. The evergreen herbs A. europaeum and H. nobilis have a life span of 13 months. Although O. acetosella is also a wintergreen species, it continuously replaces leaves during the growing season (Tessier 2004), and the average leaf longevity of this species is therefore ≈6 months in our climate. As there can be important variation in leaf longevity within species groups (Diemer 1998), our estimates of Λ are crude. Nevertheless, we suggest that even these rough estimates allow insights into the correlations between leaf longevity, leaf structure and investment in support.

quantification of biomass partitioning within the lamina and leaf shape

Representative leaves from each plant were selected for detailed analyses. In compound-leaved species, a representative leaflet was further chosen. The selected leaf lamina or leaflet lamina was divided between the mid-rib and the rest of the lamina. After determination of the fresh mass of the mid-rib and the rest of the lamina, all leaves were photocopied for area measurements.

Dry masses of the mid-rib and lamina were estimated after oven-drying at 70 °C for at least 48 h, and dry to fresh mass ratios were calculated for the leaf fractions. The partitioning in support within leaf lamina (FMR) was characterized as the ratio of mid-rib dry mass (MMR) to the sum of MMR and the rest of the lamina (ML).

The photocopied leaf images were scanned with a resolution of 300 dpi, and area, perimeter and elongation (ratio of leaf major and minor semi-axes, ɛ;Baxes 1994) were estimated separately for the selected leaves and leaflets with UTHSCSA imagetool 2·00 alpha (C. Donald Wilcox, S. Brent Dove, W. Doss McDavid and David B. Greer, Department of Dental Diagnostic Science, The University of Texas Health Science Center, San Antonio, TX, USA; http://ddsdx.uthscsa.edu). Leaf dry mass per unit area (MAW) and lamina (without mid-rib) dry mass per unit area (MA), were also calculated for the leaves sampled.

Leaf elongation characterizes the overall slenderness of the leaves. To characterize the complexity of the lamina outline, we calculated the adimensional ratio of lamina (perimeter)2/area (γ) (Sack et al. 2003b). Leaf perimeter-to-area ratio (m−1) has also been used to evaluate the degree of lamina dissection, but in contrast to γ, it is negatively correlated with leaf size (Sack et al. 2003b). For all data pooled in our data set, we observed r2 = 0·85, P < 0·001 for a correlation between perimeter/area and lamina fresh mass. Thus we prefer γ over perimeter/area ratio in our analysis.

determination of minor vein density

Leaves and leaflets sampled for detailed analyses of biomass partitioning were also used to determine the density of veins. Digital images from the lower leaf surface were made in bright field using a Nikon Eclipse E-600 microscope (Nikon Corporation, Tokyo, Japan) equipped with a Nikon Coolpix 990 digital camera. The effective magnification was 30–120× (size of sampled area, 4–60 mm2). Three to five images were taken in different positions on the same leaf. The length of secondary, tertiary and quaternary veins was measured from every image using UTHSCSA imagetool 2·00 alpha, and the average density of veins was calculated (ρV, mm mm−2). Leaf-specific mean values were further averaged to obtain a species-specific average value of vein density.

chemical analyses

Lamina and mid-rib C and N contents were determined for every selected leaf and leaflet using a Perkin Elmer series II CHNS/O Analyzer 2400 (Perkin Elmer Life and Analytical Sciences, Inc., Boston, MA, USA). Values of C and N content were averaged for all individual leaves sampled for a given species to determine species-specific means.

data analyses

We used linear and non-linear regression analyses in the form y = a log(x) + b and y = axb to test for statistical relationships between leaf characteristics. To improve the linearity and normality of these relationships, lamina fresh (MFL) and dry (MDL) mass and leaf longevity (Λ) were log-transformed before statistical analysis. All regressions were considered significant at P < 0·05. Multiple linear regressions were employed to examine the effect of leaf dry to fresh mass ratio on the fractional biomass investment in support, in leaves of varying size.

Means of different characteristics were separated among species groups (pinnate- vs parallel/palmate-veined species; ovate- vs elliptic-leaved species) using anova. In these comparisons, only species with leaf elongation (ɛ, ratio of leaf length to width) <4 were included, to avoid the bias due to species with narrow leaves that were present in some leaf shape/venation combinations. The statistical relationships between species groups (pinnate-veined vs parallel/palmate-veined; ovate vs elliptic leaves; compound vs simple leaves) were compared with covariance analyses (ancova) including all species. For non-linear relationships, the independent variable was log-transformed before the ancova. A separate-slope ancova was used first to test for the intercept differences among the groups. Whenever the interaction term (grouping variable) × covariate was statistically not significant (P > 0·05), the analysis was continued according to a common-slope model to test for the intercept differences (Sokal & Rohlf 1995).

As the species were sampled from two sites that differed in edaphic conditions, in over- and understorey dominants and in light availability, we also tested for site effect on leaf structure and biomass partitioning using anova; and on leaf structure and biomass partitioning vs log(MFL) and log(MDL) relationships using covariation analyses. According to anova, and separate-slope and common-slope ancova models, site × covariate interactions and site effects alone were not significant in any of the statistical relationships analysed (P > 0·3), suggesting that size-dependent modifications in leaf characteristics and biomass allocation between support and physiological structures occurred similarly in both sites. Thus data from both sites were pooled in the following analyses. All symbols are explained in Table 2.

Results

vein density and within-lamina support investments in leaves with different shape and venation

The average density of minor veins (ρV) varied from 0·23 to 0·91 mm mm−2 for 16 species with parallel or palmate venation, while the range of variation was 0·15–0·95 mm mm−2 for 28 species with pinnate venation. For leaves of similar elongation (ɛ < 4), ρV was larger for pinnate-veined (average ± SE = 0·62 ± 0·05) than for parallel/palmate-veined species (0·48 ± 0·07; P < 0·02). This overall difference was due to larger ρV in pinnate-veined leaves with elliptic shape (Table 3). In ovate leaves, ρV was not significantly different among pinnate- and parallel/palmate-veined species (Table 3).

Table 3.  Average (± SE) density of minor veins (ρV), mid-rib to total lamina dry mass ratio FMR = MMR/(MMR + ML), lamina dry mass per unit area (MA), lamina dry to fresh mass ratio (DF), and lamina nitrogen (NL) and carbon (CL) contents per dry mass in species with different foliage venation and shape*
Venation typeLeaf shapeρV (mm mm−2)FMR (g g−1)MA (g m−2)DF (g g−1)NL (%)CL (%)n
  • *

    Species as described in Table 1. Three ovate-leaved and two elliptic-leaved species with leaf elongation (ratio of leaf length to width) >4 were removed from the statistical comparisons (Fig. 2a). Ovate-leaved group includes species with the centroid of lamina mass located at distance a from leaf attachment <L/2, where L is total leaf length. The elliptic-leaved group consists of species with the centre of lamina mass located at aL/2. In this analysis, leaves with cordate, ovate or palmate shape were considered ‘ovate’, while leaves with elliptic, obcordate, oblong, obovate, orbicular, oval or reniform shape were considered ‘elliptic’. The species groups were compared using anova and means were separated using α Bonferroni tests. Means with the same lower case letter are not significantly different among groups (P > 0·05).

Parallel/palmateOvate0·59 ± 0·11ab0·159 ± 0·051a34·1 ± 2·1a0·210 ± 0·038a2·63 ± 0·48a41·8 ± 0·8a 7
Parallel/palmateElliptic0·39 ± 0·07a0·089 ± 0·028a33·2 ± 4·8a0·212 ± 0·032a2·8 ± 0·6a42·4 ± 0·7ab 8
PinnateOvate0·55 ± 0·08ab0·150 ± 0·016a30·2 ± 2·5a0·219 ± 0·015a2·47 ± 0·18a43·32 ± 0·41ab15
PinnateElliptic0·71 ± 0·06b0·165 ± 0·037a30·3 ± 3·7a0·277 ± 0·035a2·39 ± 0·20a44·1 ± 0·6b10

The fraction of lamina biomass in the mid-rib [FMR = MMR/(ML + MMR)] varied from 0·017 to 0·39 g g−1 for parallel/palmate-veined species, and from 0·048 to 0·40 g g−1 for pinnate-veined species. The mid-rib dry mass fraction did not depend on leaf venation and shape (Table 3).

Vein density scaled negatively with leaf elongation in pinnate-veined species (Fig. 2a), while ρV was negatively associated with (perimeter)2/area (γ) only in pinnate-veined elliptic leaves (r2 = 0·45, P < 0·05 for a correlation in the form y = axb), but not for other venation/shape combinations (P > 0·1). The fraction of lamina mass in the mid-rib was positively related to lamina elongation (ɛ) (r2 = 0·82, P < 0·001) and γ (r2 = 0·65, P < 0·01) in pinnate-veined elliptic leaves. For other venation/shape combinations, r2 < 0·20, P > 0·3 for both ɛ and γ.

Figure 2.

Density of minor leaf veins (ρV, second-, third- and fourth-order veins) in relation to leaf elongation (adimensional ratio of leaf major and minor semi-axes) (a); and mid-rib to whole lamina (mid-rib + rest of lamina) dry mass ratio, g g−1[FMR = MMR/(ML + MMR)] (b) in different leaf types. The ‘parallel-veined’ species group consists of species with parallel and palmate venation (Fig. 1; Table 1). ‘Ovate’ leaf shape includes species with the lamina mass centre, x(c), located at distance from leaf attachment a < L/2, where L is total leaf length (Fig. 1). ‘Elliptic’ leaf shape includes leaves with the lamina mass centre located at distance a ≥ L/2. Leaves with cordate, ovate or palmate shape were considered ‘ovate’; leaves with elliptic, obcordate, oblong, obovate, orbicular, oval or reniform shape were considered ‘elliptic’ (Fig. 1; Tables 1 and 3). The x-axis was log-transformed to linearize the relationships and data were fitted by linear regressions. For parallel/ovate species groups, n = 7; for parallel/elliptic, n = 8; for pinnate/ovate, n = 18; for pinnate/elliptic, n = 11.

Negative relationships were observed between ρV and lamina mass in the mid-rib for pinnate-leaved species, and for parallel-veined ovate-leaved species (Fig. 2b). According to ancova (venation type as main effect, log(ρV) as covariate), pinnate-leaved species had a larger ρV at a common fractional biomass investment in the mid-rib (P < 0·001). This effect resulted from lower ρV in parallel-veined elliptic leaves than in ovate and elliptic pinnate-veined leaves (P < 0·001).

leaf structure and chemistry in leaves of varying shape and venation

Lamina dry mass per unit area (MA), dry to fresh mass ratio (DF), lamina N content per dry mass (NL, Table 3), and mid-rib N content per dry mass (NMR, P = 0·16) were not different among leaf shape/venation combinations. However, lamina carbon content per dry mass (CL) was significantly larger for pinnate-veined elliptic leaves than for parallel/palmate-veined ovate leaves (Table 3), and the average lamina carbon content of 43·60 ± 0·33% of all pinnate-veined species pooled was also significantly larger (P < 0·02) than that of parallel/palmate-veined species pooled (42·1 ± 0·5%).

Lamina dry to fresh mass ratio was positively correlated with MA when all data were pooled (r2 = 0·18, P < 0·001), but among species groups this relationship was significant only for parallel/palmate-veined elliptic leaves (r2 = 0·63, P < 0·02). DF was negatively correlated with NL (Fig. 3a, r2 = 0·39, P < 0·001 for all data pooled), and positively correlated with CL (r2 = 0·31, P < 0·001 for all data pooled), while neither NL (r2 = 0·01, P > 0·6) nor CL (r2 = 0·07, P > 0·09) was correlated with MA. Neither the slopes (P > 0·09) nor the intercepts (P > 0·2) of NL vs log(DF) (P > 0·4) and CL vs log(DF) relationships were significantly different among the leaf venation/shape pairs.

Figure 3.

Correlations of lamina nitrogen content per dry mass (NL; a), ρV (b) and FMR (c) with lamina dry to fresh mass ratio in species with different venation and shape (symbols, species grouping and sample size as in Fig. 2; see Table 1 for species). Data were fitted by non-linear regressions in the form y = a + b log(x).

Lamina N content was 1·8-fold larger than that of mid-ribs (P < 0·001 according to a paired samples t-test). Lamina carbon content of 43·09 ± 0·30% was also larger than that of the mid-rib of 40·06 ± 0·43% (P < 0·001 for differences among means). Lamina and mid-rib N contents were strongly correlated (r2 = 0·75, P < 0·001). This relationship was not different among leaf venation/shape pairs (P > 0·1 for slope differences and P > 0·7 for intercept differences). Lamina and mid-rib carbon contents were also positively correlated (r2 = 0·25, P < 0·01), and this relationship was not affected by leaf shape or venation (P > 0·5 for slope differences; P > 0·09 for intercept differences).

vein density and lamina support in relation to leaf structure and carbon content

Vein density was positively associated with DF in ovate-leaved species (Fig. 3b), and for all data pooled (r2 = 0·43, P < 0·001). Neither the slopes (P > 0·5) nor the intercepts (P > 0·1) of ρV vs log(DF) relationships differed among the species groups. The fraction of lamina biomass in the mid-rib was negatively correlated with DF in all cases (Fig. 3c, r2 = 0·53, P < 0·001 for all data pooled). The slope of FMR vs log(DF) was not different among species groups (P > 0·2), but parallel-veined elliptic leaves had significantly lower FMR at a common DF than pinnate-veined ovate leaves (P < 0·03 for intercept differences) and elliptic leaves (P < 0·002). The relationships of ρV (r2 < 0·38, P > 0·1 for various venation/shape groups, r2 = 0·01, P > 0·4 for all data pooled) and FMR (r2 < 0·2, P > 0·2 for various venation/shape groups, r2 = 0·03, P > 0·3 for all data pooled) with MA were not significant.

For all data pooled, lamina carbon content was positively correlated with ρV (r2 = 0·18), but negatively correlated with FMR (r2 = 0·17, P < 0·02 for both).

leaf structure and investment in support in relation to leaf longevity (Λ)

For all data pooled, leaf longevity was positively correlated with leaf dry to fresh mass ratio (Fig. 4a) and lamina carbon content per dry mass (r2 = 0·27, P < 0·001 for log–log-transformed data), and negatively with lamina nitrogen content per dry mass (r2 = 0·18, P < 0·001 for log–log-transformed data), demonstrating that leaves with greater longevity had a more robust design. Vein density was not associated with Λ (r2 = 0·06, P > 0·16), but the fraction of leaf biomass in the mid-rib was negatively associated with Λ (Fig. 4b). As Λ was non-uniformly distributed among venation and shape classes, correlations of structural and chemical characteristics with Λ were not tested within species groups.

Figure 4.

Lamina dry to fresh mass ratio (a) and fraction of mid-rib in lamina (b) depending on leaf life span in leaves of different shape and venation (symbols as in Fig. 2). Insets demonstrate correlations for log–log-transformed data. (a) O.a.=Oxalis acetosella, a wintergreen species having soft leaves with high water content (r2 = 0·44 without this observation). (b) R.c. =Ranunculus cassubicus, which has palmately veined leaves with a non-pronounced mid-rib (r2 = 0·62 without this observation).

scaling of lamina support costs with lamina size

The fraction of lamina mass in the mid-rib was positively associated with MFL (Fig. 5, r2 = 0·48 for all data pooled and r2 = 0·57, P < 0·001 without T. officinale, which had strongly dissected leaves and possessed the largest FMR at a common leaf mass) and lamina dry mass (r2 = 0·40 with; r2 = 0·48 without T. officinale, P < 0·001 for both). The slopes of FMR vs log(MFL) were not significantly different among species groups (P > 0·07), but at a common log(MFL), FMR was lower in parallel/palmate-veined elliptic leaves than in pinnate-veined elliptic leaves (Fig. 5, P < 0·03 without; P < 0·04 with T. officinale). Vein density was not correlated with either lamina fresh mass (r2 = 0·02, P > 0·4) or dry mass (r2 = 0·00, P > 0·8).

Figure 5.

Fraction of mid-rib in lamina (FMR) in relation to lamina + mid-rib fresh mass in leaves of different shape and venation (symbols as in Fig. 2). T.o., an outlying observation in Taraxacum officinale, which has strongly pinnatisect leaves. Punctuated line shows the regression fitted to the data without this outlying observation.

In multiple regression analyses of FMR vs log(MFL) and log(DF), log(DF) was generally negatively associated with FMR (P < 0·02 for log(DF) for all venation/shape pairs, except for parallel-veined ovate leaves, P > 0·1), and inclusion of DF into the regressions increased the variance explained (r2 = 0·95 for parallel-veined ovate; r2 = 0·83 for parallel-veined elliptic; r2 = 0·81 for pinnate-veined ovate; r2 = 0·70 for pinnate-veined elliptic leaves). Although log(DF) was non-significant in multiple regressions for parallel-veined ovate leaves, FMR scaled positively with log(MFL) and negatively with log(DF) for all data pooled (r2 = 0·67, P < 0·001 for both). Analogously, FMR scaled positively with both log(MFL) and leaf longevity (r2 = 0·64, P < 0·001 for both).

Discussion

vein density in leaves of different shape and venation

We hypothesized that the density of minor veins (ρV) is larger for pinnate-veined species, because the major veins are less uniformly distributed within the leaf in pinnate-veined than in parallel- and palmate-veined species (Fig. 1). All data collectively supported this hypothesis (Table 3). Vein density was further strongly associated with lamina carbon content (CL). As leaf carbon contents scale with the content of carbon-rich chemicals such as lignin (Niinemets et al. 1999; Poorter 1994), CL provides an alternative estimate of within-lamina investments in support, including minor veins and cell walls. This positive correlation, and larger lamina carbon contents in pinnate-veined leaves, suggest an overall greater biomass investment in within-leaf support (minor veins and cell walls) in these leaves (Table 3).

However, the difference in ρV and CL between leaves with different venation types mainly reflected the difference among leaves with the centre of lamina mass located further than half the leaf length (‘elliptic’ leaves, Fig. 1), while the average values of ρV and CL did not differ among pinnate- and parallel/palmate-veined ovate leaves. This may reflect the overall greater support requirements of mechanically less efficient elliptic leaves.

Modifications in venation architecture also affect leaf hydraulic characteristics, and can be driven partly by requirements for efficient water conduction. Strong positive correlations have been observed between vein density and leaf hydraulic conductance (Sack & Frole 2006). Thus optimization of leaf hydraulic architecture can provide an alternative explanation for the observed differences in vein density. Because the hydraulic conductance of veins is not infinite, significant water potential gradients develop from the lamina base towards the tip, resulting in particularly low lamina water potentials at the apical lamina parts (Zwieniecki et al. 2002). In elliptic leaves, in which the tapering mid-rib supplies a disproportionately large part of the lamina area, hydraulic limitations in apical portions of pinnately veined leaves are expected to be especially severe. Minor veins constitute often the largest fraction of total leaf hydraulic resistance (Sack et al. 2004, 2005), and increases in the number of parallel vascular paths through enhanced vein density may provide an important compensatory way to reduce water potential gradients along the lamina. Overall, the greater vein density in pinnate-veined leaves agrees with the evidence that whole-leaf hydraulic conductance is determined more strongly by the conductance of minor veins in pinnate-veined than in palmate-veined leaves (Sack et al. 2003b, 2004).

As leaf dissection, lobation and serration patterns follow the distribution of secondary and tertiary veins, increases in the complexity of leaf perimeter have been associated with increases in vein density (Sisóet al. 2001). In our study, the ratio lamina (perimeter)2/area was associated with vein density only in pinnate-veined elliptic leaves, but this correlation was negative. Although, in our study, only a few species had strongly dissected laminas, this contrasting pattern suggests that the effects of lamina dissection are superimposed by overall modifications of venation architecture and lamina shape.

For pinnate-veined leaves, we further observed a decrease in vein density with leaf elongation (Fig. 2a), indicating that with decreasing the average distance of leaf margins from the mid-rib, the requirement of minor veins for mechanical support or water conduction is lower. These data collectively indicate a significant interspecific variation in vein density that is associated with venation architecture, lamina shape and elongation, and has important implications for lamina hydraulic efficiency and mechanical support.

investment in the mid-rib in relation to leaf size, vein density and venation architecture

Our study highlights a large fractional investment of lamina biomass in the mid-rib (FMR), often 0·1–0·2 g g−1 in medium-sized leaves and up to 0·3–0·4 g g−1 in the largest dissected leaves. As larger leaves have effectively longer lever arms, biomechanical considerations for leaf self-support suggest that larger leaves should have a relatively larger fraction of biomass in the mid-rib (Givnish 1979, 1986). However, experimental data to test this biomechanical principle have been scarce, especially for a large set of species (for review see Niinemets & Sack 2006). Our study does indicate a strong scaling of FMR with leaf size across a range of species with contrasting shape and venation (Fig. 5).

Although we hypothesized that elliptic leaves with effectively longer lever arms have greater FMR, this hypothesis was not supported by the data (Table 3; Fig. 5). In fact, we observed a broad trade-off between mid-rib investments and vein density (Fig. 2b). This trade-off highlights that lamina support requirements can be met by either enhanced biomass investments in the mid-rib or in minor veins. This is possible as different combinations of total support biomass partitioning between minor and major veins can yield similar lamina flexural rigidity. Unexpectedly, elliptic leaves were not significantly distinguished from ovate leaves according to this relationship (Fig. 2b), possibly indicating a greater safety margin against mechanical failure in ovate leaves.

In our study, differences in venation architecture were a more important determinant of differences in support investments than leaf shape per se. Theoretical studies suggest that, for a common biomass investment in major veins, palmate- and pinnate-veined leaves are hydraulically and mechanically more efficient than pinnate-veined leaves (Jeje 1985; Kull & Herbig 1995). Although such theoretical studies provide important insight into the optimality of venation architecture, such simulations are valid only for specific leaf size/shape combinations, and need to consider higher-order veins as well.

We observed that parallel/palmate-veined elliptic leaves had a lower vein density at common FMR than ovate and elliptic pinnate-veined leaves (Fig. 2a). As in pinnate-leaved species, the mid-rib also appears to bear the major bending stresses in elongated parallel- and palmate-veined species (Moulia et al. 1994; Moulia & Fournier 1997). However, in our study, parallel/palmate-veined species were generally also less extended than pinnate-leaved species (Fig. 2a). Given that the conductance of veins is inversely proportional to the fourth power of their diameter (for review see Tyree & Jarvis 1982), a central vein and high density of minor veins is not apparently an optimal strategy for wide leaves. In fact, as the distance of the lamina margin from the central support element increases, leaf hydraulic characteristics can be optimized if the biomass invested in major veins is distributed between several veins radiating from the base, instead of a single central vein. Thus the overall greater density of major veins in less elongated palmate/parallel-veined species is probably another part of the general trade-off between the investments in minor and major veins observed for all data pooled (Fig. 2b).

lamina chemistry, structure, longevity and investments in support

Previous analysis suggests that larger leaves have a significant disadvantage because of disproportionately larger support costs. Although support tissues also contain chlorophyll and photosynthetic enzymes (Niinemets 1999a; Hibberd & Quick 2001), the photosynthetic activity of support structures is low (Niinemets 1999a). Significantly lower N contents of mid-ribs relative to lamina found in our study also support this suggestion.

Extending a larger fraction of leaf area further from stem and leaf attachment can compensate for this drawback by decreased self-shading and enhanced light-capture efficiency (Pearcy & Yang 1998). In addition, we observed a negative correlation between the fraction of lamina biomass in mid-rib and lamina dry to fresh mass ratio (DF, Fig. 3c). As multiple regression analyses demonstrated, DF affected the fractional biomass investment in the mid-rib independently of leaf size. As increases in DF result mainly from a greater fraction of cell walls, DF is negatively correlated with leaf physiological activity (Garnier & Laurent 1994). The uniform negative correlation between DF and lamina nitrogen content (Fig. 3a) further underscores this pattern, and suggests that a larger mid-rib mass fraction is combined with greater physiological activity of the rest of the lamina. In addition, DF was positively, and the biomass fraction in the mid-rib negatively, associated with leaf longevity (Fig. 4). These relationships are consistent with the large body of information of lower functional activity of long-lasting leaves (Reich et al. 1992, 1999; Wright et al. 2004). The negative relationship between longevity and FMR suggests that lifetime support tissue investment is larger in short-lived leaves. This pattern is apparently counterintuitive, but emphasizes the large support biomass cost of high functional leaf activity.

Turgor (the pressure exerted by cellular water on cell walls) can contribute strongly to leaf support in species with thin cell walls and high leaf water content (Niklas 1986, 1989). As species with lower DF are expected to rely more strongly on turgor for mechanical support, we suggested that, among the leaves of common mass, area and shape, dry mass investment in the mid-rib is lower for leaves with lower DF. However, the negative correlation between DF and the biomass fraction in the mid-rib suggests that reliance on turgor for mechanical stability does not necessarily mean lower biomass investments in support. Given that leaves with high physiological activity also have greater water loss via transpiration than plants with lower physiological activity (for correlations between leaf structure and stomatal conductance see Schulze et al. 1994; Reich et al. 1997), rapid water loss must be accompanied by rapid water supply for turgor maintenance. Although plants with higher water content can rely more heavily on turgor for mechanical stability, our data suggest that mechanical support through turgor is not necessarily a cheap solution in terms of biomass investment in support.

The negative relationship between FMR and DF is surprising at first glance, but may provide a major means by which large and structurally expensive leaves offset the high cost of leaf construction. In this regard, it is interesting that there are well known positive relationships between site nutrient availability and average leaf size of dominant species (Givnish 1978). Our data suggest that this general worldwide relationship may reflect an enhanced within-leaf support cost of large-leaved species, and accordingly inherent requirements for high soil nutrient availabilities to achieve high foliage physiological capacities that allow the leaves to pay back the costs of their construction.

Previously, it has been demonstrated that large-leaved species generally have relatively lower stem xylem cross-sectional areas for a common leaf area supported (Pickup et al. 2005). If so, this could provide an additional strategy to compensate for high within-leaf support costs. Although we investigated mostly herbs in our study, it is interesting that in several of the large-leaved species studied, the petioles and stems are hollow (e.g. Arctium tomentosum, C. pepo, H. sosnowskyi, L. wilsoniana, Rumex crispus, Tussilago farfara), suggesting that there may be a general trade-off between within-leaf and within-plant support investments (Niinemets et al. 2006b).

Conclusions

There is extensive variation in within-plant support investments among plant species (Valladares et al. 2002; Niinemets et al. 2006b). Our study further highlights a dramatic, more than eightfold variation in within-leaf support costs among leaves of different shape, size and venation, and an important trade-off between investments in minor and major veins, suggesting that leaves of different size and shape are characterized by different combinations of mid-rib investment and minor vein density. This quantitative gradient in investments in minor vs major veins has important implications for leaf physiological activity and hydraulic efficiency. While there appear to be fundamental scaling relationships among leaf physiological and structural characteristics (Wright et al. 2004), we suggest that, among leaves of common integrated characteristics such as leaf N content and DM per unit area, partitioning of leaf hydraulic conductance between minor and major veins and overall investments in support may lead to a gradation of foliar physiological activity at specific points along the continuous spectrum of leaf functioning. This spectrum of vein density vs biomass investment in mid-rib may have important implications for niche differentiation of coexisting species.

Acknowledgements

We thank the Estonian Science Foundation (grant 4584), the Estonian Ministry of Education and Science (grant 0182468As03), and the Estonian Academy of Sciences for financial support.

Ancillary