Relating root structure and anatomy to whole-plant functioning in 14 herbaceous Mediterranean species

Authors


Author for correspondence: Catherine Roumet Tel: +33 4 67 61 32 38 Fax: +33 4 67 41 21 38 21 Email: catherine.roumet@cefe.cnrs.fr

Summary

  • • This study investigated the relationships between root structure and anatomy and whole-plant functioning in herbaceous species.
  • • Fourteen annual and perennial species representative of a Mediterranean old-field succession were grown in monocultures in a common-garden experiment. Whole-plant functioning was assessed by inherent relative growth rate (RGRmax), measured in standardized conditions, and maximum height (Hmax). Root tissue density (TMDr), considered as a major component of root structure, was measured on roots harvested within in-growth cores. Anatomical characteristics were analysed on cross-sectional areas (CSA).
  • • TMDr was correlated positively with Hmax and negatively with RGRmax. Root CSA explained interspecific variation in Hmax but not that in TMDr and RGRmax. Root xylem CSA and xylem proportion in root CSA were positively correlated with TMDr and Hmax and negatively with RGRmax. Mean xylem vessel CSA did not account for variations in TMDr, Hmax and RGRmax.
  • • These results suggested that RGRmax and Hmax are constrained by opposite root structural and anatomical traits, which have potential links with hydraulic conductance, support and longevity.

Introduction

Tissue mass density (TMD), defined as the ratio of dry mass to fresh volume, has become an important trait in comparative plant ecology, in that it belongs to a syndrome of traits involved in the acquisition–conservation trade-off plants have to perform (Wilson et al., 1999; Ryser & Urbas, 2000; Craine et al., 2001; Garnier et al., 2001; Preston et al., 2006). Within organs, TMD is the expression of the relative investment of dry matter between different types of tissue with functions varying from support to growth (Garnier & Laurent, 1994; Ryser, 1996; Ryser, 1998; Wahl & Ryser, 2000). At the whole-plant level, TMD arises as an inherent constraint that prevents the simultaneous maximization of resource acquisition and conservation (Ryser & Eek, 2000). Interspecific variation in TMD reflects variations both in plant traits involved in resource capture, and in anatomical traits involved in support or growth. As an example, plant height, which influences plant support and light interception in herbaceous communities, covaries with stem TMD (Huston & Smith, 1987; Grime et al., 1988; Givnish, 1995; Falster & Westoby, 2005) and with the proportion of sclerified tissues in stems (Enquist et al., 1999). Inherent relative growth rate (RGRmax), which reflects resource acquisition (Grime, 1979; Chapin, 1980), covaries with shoot TMD. Fast growers show low-density tissues, expansion of which reduces the dry matter cost (Ryser & Lambers, 1995; Reich et al., 1998; Ryser, 1998; Wright & Westoby, 1999). They are characterized by leaves in which a high fraction of volume is occupied by mesophyll, the assimilatory tissue (Garnier & Laurent, 1994; van Arendonk & Poorter, 1994). On the other hand, slow growers produce high-density tissues, which favour both organ persistence and conservation of resources (Gleeson & Tilman, 1994; Reich et al., 1995; Bazzaz, 1996; Wahl & Ryser, 2000; Cooley et al., 2004; Garnier et al., 2004). They exhibit leaves with a high volume of sclerified tissue (Garnier & Laurent, 1994; Van Arendonk & Poorter, 1994; Cornelissen et al., 1996; Castro-Diez et al., 2000) and stem with a high proportion of xylem in stem cross-sectional area (CSA) (Castro-Diez et al., 1998). While the relationship between TMD, anatomy and plant functioning is widely admitted for shoots, there is little information about such trade-offs at the root level (Eissenstat, 2000).

Roots are physically linked to stems and their shape is constrained by mechanical laws and allometric scaling (Enquist, 2003). Their structure and anatomy have evolved in response to distinct selection pressures, such as anchorage or nutrient and water uptake, in relation to the resource cost of root production and maintenance (Raven & Edwards, 2001). Among structural properties, TMD of roots (TMDr) is considered one of the major components of root structure discriminating species: it shows contrasting differences between life spans and botanical families (Craine et al., 2001; Roumet et al., 2006); within perennial grasses it is related to root longevity (Ryser, 1996) and to RGRmax (Wahl & Ryser, 2000); finally, it varies along fertility and disturbance gradients (Craine et al., 2001). Root anatomical traits are also intimately involved in whole-plant functioning. A negative correlation exists between hydraulic conductivity and root diameter or cortex width in herbaceous and woody species (Rieger & Litvin, 1999). In 19 perennial grasses, anatomical structures related to hydraulic conductance, such as xylem vessel area and the proportion of xylem, are associated with plant height at maturity (Hmax), while those contributing to root robustness, such as stele CSA, the proportion of stele, and the proportion of cell wall in the stele, constrain RGRmax (Wahl & Ryser, 2000). Inherent differences in root anatomy and TMDr might thus reflect complex trade-offs in whole-plant functioning. These conclusions require to be tested on a wider range of species. Species comparisons provide a means of testing general trade-offs in root form and function (Peterson, 1992).

The aim of this study is to investigate whether two key traits in whole-plant functioning, RGRmax and Hmax, are related to TMDr and root anatomy. To this end, we focused on 14 herbaceous species representative of a Mediterranean old-field succession; they included monocots and dicots, annuals and perennials belonging to several families. This ecological context offers the opportunity to work with species exhibiting a wide range of variation in plant traits; previous studies have shown that these species differed 1.9-fold in RGRmax measured in controlled conditions (Vile et al., 2006a) and 15-fold in Hmax in natura (Vile et al., 2006b). In the present study, the 14 species were grown in monocultures in a common-garden experiment. The in-growth core method (Eissenstat, 1991; Neill, 1992) was used to produce comparable root material (roots produced within the same period). Anatomical characteristics were studied to determine which features underlie interspecific variations in TMDr, RGRmax and/or Hmax.

Materials and Methods

Species and growth conditions

Fourteen herbaceous species were selected among the most abundant ones occurring in French Mediterranean old-field successions (Escarréet al., 1983; Garnier et al., 2004) (Table 1). These species belong to different life spans (seven annuals, two biennials and five perennials) and to eight botanical families (Table 1). They are representative of three stages of succession following vineyard abandonment: early (2–6 yr), intermediate (7–15 yr) and advanced (15–45 yr) (Garnier et al., 2004). The intermediate stage of succession exhibited all life spans, whereas the early and advanced stages were represented by only annuals and perennials, respectively (Table 1). Two families, Poaceae and Asteraceae, were represented in each successional stage; it is thus possible to highlight shifts in a similar direction for different lineages as succession proceeds.

Table 1.  Species studied, assigned to a successional stage based on their abundance and usual position in Mediterranean old-field successions (Garnier et al., 2004)
SpeciesFamilyLife spanSuccessional stage
  1. Nomenclature follows Tutin et al. (1968–80).

Arenaria serpillyfolia (L.)CaryophyllaceaeAnnualEarly
Bromus madritensis (L.)PoaceaeAnnualEarly
Crepis foetida (L.)AsteraceaeAnnualEarly
Geranium rotundifolium (L.)GeraniaceaeAnnualEarly
Veronica persica (Poir.)ScrophulariaceaeAnnualEarly
Trifolium angustifolium (L.)FabaceaeAnnualIntermediate
Tordylium maximum (L.)ApiaceaeAnnualIntermediate
Daucus carota (L.)ApiaceaeBiannualIntermediate
Picris hieracioides (L.)AsteraceaeBiannualIntermediate
Calamintha nepeta (L.) SaviLamiaceaePerennialIntermediate
Dactylis glomerata (L.)PoaceaePerennialIntermediate
Bromus erectus (Huds.)PoaceaePerennialAdvanced
Brachypodium phoenicoides (L.)
Roemer & SchultesPoaceaePerennialAdvanced
Inula conyza (D.C.)AsteraceaePerennialAdvanced

Each species was grown in monoculture from November 2003 to July 2005 in a common-garden experiment conducted at the Centre d’Ecologie Fonctionnelle et Evolutive (CEFE-CNRS) in Montpellier, France (43°59′ N, 3°51′ E, 60 m asl). The climate is Mediterranean subhumid (Daget, 1977) with cool to cold winters, marked summer drought, frequent frosts in winter, and unpredictability of precipitation in time and amount. During the experiment, the mean annual rainfall and temperature were 285 mm and 14.5°C, respectively (CEFE meteorological station). Soil (0–20 cm) was sandy loam with pH = 7.3 ± 0.2.

Seeds or ramets of each species were collected in the successional sere described by Garnier et al. (2004) and grown for 2 months in glasshouse. When seedlings were large enough, they were transplanted in the common-garden experiment. Four replicate monocultures of each species (1.2 × 1.2-m plots) were established in autumn 2003 at a density of 100 plants m−2. Within plots, seedlings were 10 cm apart. After establishment, no water or nutrients were added. Plots were weeded regularly to maintain density and monoculture status. In autumn 2004, annuals and biennials, collected and grown as above, were retransplanted on the same plots.

Root measurements

Root sampling  Roots were sampled twice, in March and May 2005, when perennials had been growing since autumn 2003 and annuals and biennials since autumn 2004. Within each plot a soil core (5 cm diameter, 20 cm depth) was taken randomly between four equidistant plants, except border plants, in both January and March 2005, at different places within each plot. Each hole was refilled with sieved soil, free of roots. At 8 wk later a soil core (3 cm diameter, 20 cm depth) was harvested inside the previous one. Roots were carefully washed free of soil in water. The root material harvested consisted of several root fragments with a known maximum age of 8 wk. The developmental order was impossible to determine as root tips were not always present. Roots from the March harvest were used immediately for root anatomy measurements. Roots from the May harvest were stored in ethanol 50% (v/v) until morphological measurements.

Root tissue mass density (TMDr)  All root fragments produced within the March in-growth core were used for TMDr determination. Roots were stained with methylene blue (5 g l−1) to increase contrast during scanning. They were rinsed, spread out in water onto a mesh tray, and finally transferred onto a transparent acetate sheet and scanned at 400 dpi. A digital image-analysing system (whinrhizo ver. 2003b, Regent Instruments Inc., Quebec, Canada) was used to determine root volume assigned to 12 diameter classes ranging from 0.05 to 0.6 mm, the maximal diameter recorded among root fragments. Within each diameter class the root volume was calculated using the real diameter. Roots were oven-dried for 48 h at 60°C and weighed. The TMDr was calculated as the ratio of root dry mass to total root volume (cumulated volume of the different diameter classes).

Root anatomy  Root fragments harvested in the March in-growth core were immediately dipped in a fixative containing 0.1 m phosphate buffer pH 7, 1% glutaraldehyde, 2% paraformaldehyde and 1% caffeine, and observed under a binocular microscope at ×50 magnification. First-order roots (terminal roots with growing root tips; ‘functional method’sensu Fitter, 1987) were selected. Sections (4 mm long) were taken 10 mm above the root tip, in the elongation zone. This procedure controlled sample uniformity, as first-order roots are most likely to serve similar functions of water and nutrient uptake (Rieger & Litvin, 1999; Eissenstat, 2000). Sections were dipped in fixative, placed under vacuum for 15 min, and incubated overnight at 4°C. Samples were dehydrated by dipping them in 70% ethanol (2 × 1 h) before being embedded into 6% (w/v) agar. Embedded samples were then moved through an ethanol dehydration series: 24 h in 70% ethanol, 2 × 2 h in 95% ethanol, and finally rinsed and stored in absolute ethanol at 4°C. Dehydrated samples were embedded in Technovit 7100 (Kulzer, Wehrheim, Germany) and cross-sections (3 µm thick) were obtained using a microtome (RM 2255 Leica, Bensheim, Germany). Specimens were double-stained with Periodic Acid Schiff reagent combined with protein-specific naphthol blue-black (Fisher, 1968), which stains starch reserves and cell walls pink, and proteins dark blue. To contrast the xylem structure, specimens were stained with Toluidine Blue, which stains lignin blue-green, and cellulose purple (Feder & O’Brien, 1968). No cambial activity and no specific reserves were observed in any root cross-sections. Sections were examined under the light microscope (DMXRA, Leica). Image acquisition and analyses as well as anatomical measurements were made using openlab software (Improvision, Lexington, MA, USA). On each specimen, cross-sectional areas (CSA) were assessed by tracing the outlines of structures with cursor. The CSA of the whole section, total xylem, cortex and rhizodermis were measured. Up to 10 lumens of xylem vessels, randomly sampled, were measured per specimen to determine the mean xylem vessels CSA. We focused our analyses on the mean xylem vessels CSA, as this feature is more important for hydraulic conductance than the number of vessels (McCully & Canny, 1988; Wahl & Ryser, 2000). Based on these measurements, the proportion of xylem CSA per total root CSA was calculated. The number of root cross-sections observed per species is shown in Table 2.

Table 2.  Structural, growth and anatomical characteristics for the 14 species studied
SpeciesTMDr (mg cm−3)RGRmax (g g−1 d−1) H max (cm) n Root CSA (µm2)Rhizodermis + cortex CSA (µm2)Xylem CSA (µm2)Mean xylem vessel CSA (µm2)Proportion of xylem in root CSA
  1. Mean values ± SE; n, number of anatomical section. TMDr, tissue mass density of roots; RGRmax, inherent relative growth rate; Hmax, height at maturity; CSA: cross-sectional area.

Arenaria serpillyfolia  68 ± 40.266 ± 0.00413 ± 1 622380 ± 469120071 ± 4319 376 ± 80 22 ± 40.017 ± 0.001
Bromus madritensis 104 ± 60.261 ± 0.00445 ± 4 640140 ± 947335928 ± 124072307 ± 687 42 ± 100.053 ± 0.009
Crepis foetida  63 ± 50.334 ± 0.00574 ± 11059425 ± 1095153937 ± 99501253 ± 299 87 ± 160.022 ± 0.001
Geranium rotundifolium  76 ± 60.271 ± 0.00531 ± 1 921413 ± 223017472 ± 1770 817 ± 168 27 ± 40.041 ± 0.008
Veronica persica  58 ± 50.334 ± 0.00412 ± 11133933 ± 476532036 ± 4614 480 ± 77 42 ± 50.014 ± 0.001
Trifolium angustifolium 106 ± 40.213 ± 0.00843 ± 2 621383 ± 322120381 ± 38371002 ± 104 20 ± 40.054 ± 0.002
Tordylium maximum  87 ± 40.292 ± 0.00264 ± 31038218 ± 595232001 ± 46101160 ± 227134 ± 160.030 ± 0.004
Daucus carota  76 ± 80.268 ± 0.00573 ± 2 632112 ± 688927763 ± 6855 994 ± 292 63 ± 90.035 ± 0.005
Picris hieracioides  73 ± 30.291 ± 0.00660 ± 31044739 ± 577640666 ± 5603 635 ± 78 51 ± 50.015 ± 0.001
Calamintha nepeta  90 ± 130.268 ± 0.00553 ± 21123514 ± 259421705 ± 2356 884 ± 276 25 ± 30.034 ± 0.009
Dactylis glomerata 130 ± 40.244 ± 0.00673 ± 4 860913 ± 1039753747 ± 94314878 ± 953 26 ± 20.074 ± 0.006
Bromus erectus 166 ± 210.174 ± 0.00486 ± 31047611 ± 694839300 ± 62575039 ± 631 32 ± 40.100 ± 0.028
Brachypodium phoenicoides 164 ± 110.182 ± 0.00394 ± 3 634320 ± 933226951 ± 69705294 ± 1481 38 ± 50.161 ± 0.028
Inula conyza  87 ± 30.256 ± 0.00495 ± 71079191 ± 1669475558 ± 159963633 ± 164 36 ± 40.050 ± 0.002

Whole-plant measurements

Plant height  Plant height was measured monthly from November 2004 to July 2005, on four plants per plot. The Hmax was the maximum height reached and represented the reproductive height in our experiment.

Inherent relative growth rate  The RGRmax was determined at the seedling stage under controlled conditions; taken from Vile et al. (2006b) for 13 of the species and determined using an identical protocol for Inula conyza.

Data analyses

Pearson's correlation and linear regressions were performed on pooled data for the 14 species to relate growth, structural and anatomical features. Variables were loge-transformed to meet the assumptions of normality as needed. Multiple linear regression analyses were used to assess the dependencies of RGRmax and Hmax on TMDr and unrelated anatomical features. A one-way anova was performed to test for differences among life spans. All statistical analyses were performed using sas (SAS Institute, Cary, NC, USA). Because biennials behave as annuals in this experiment, they have been considered as annuals in the one-way analysis.

Results

Interspecific variations in TMDr scale those in RGRmax and Hmax

A 2.9-fold variation in TMDr was observed among the species, from 58 mg cm−3 for the annual dicot Veronica persica to 166 mg cm−3 for the perennial monocot Bromus erectus (Table 2). Interspecific variation in RGRmax was lower than that in TMDr and ranged from 0.174 for B. erectus to 0.334 g g−1 d−1 for V. persica and Crepis foetida (Table 2). The Hmax was also highly contrasted between species, as shown by the eightfold variation between V. persica and the perennial dicot Inula conyza (Table 2). Perennials had denser root tissues (anova, F = 11.4, P < 0.01), and exhibited a lower RGRmax (F = 6.44, P < 0.05) and a higher Hmax (F = 7.76, P < 0.05) than short life-span species (Fig. 1a,b). The RGRmax decreased with species successional status, while TMDr and Hmax increased (Table 1; Fig. 1a,b). The same patterns were observed within the Poaceae for these three traits and within the Asteraceae for TMDr and RGRmax (Table 1).

Figure 1.

Relationships between root tissue mass density (TMDr) and (a) relative growth rate (RGRmax); (b) height at maturity (Hmax) for 14 herbaceous species. Solid lines denote significant relationships. Each point represents a species: annuals (open symbols), perennials (closed symbols); species from early (○), intermediate (▿) and advanced (□) successional stages.

Figure 1 shows that TMDr, RGRmax and Hmax follow a continuous pattern of covariation among species. TMDr was strongly and negatively correlated with RGRmax (r = –0.91, P < 0.001; Fig. 1a) and positively correlated with Hmax (r = 0.58, P < 0.05; Fig. 1b). Thus species exhibiting a fast growth rate in a controlled experiment produce root tissues of low density under common-garden conditions. Hmax and RGRmax were, however, not significantly correlated (r = –0.44, P = 0.11). The relationships between TMDr, Hmax and RGRmax point out the importance of TMDr for whole-plant functioning.

Anatomical features contribute diversely to root CSA

A 3.7-fold variation in root CSA was observed among the species from 21 383 µm2 for the annual Trifolium angustifolium to 79 191 µm2 for the perennial I. conyza (Table 2). Interspecific variations in xylem CSA were larger than those of root CSA and ranged from 376 µm2 for the annual dicot Arenaria serpillyfolia to 5294 µm2 for the perennial monocot B. phoenicoides (Table 2). An 11-fold variation in xylem CSA as a proportion of root CSA was found: xylem CSA represented 1.4% of the root CSA in the annual dicot V. persica and 16% in the perennial monocot B. phoenicoides. Mean CSA of xylem vessels ranged from 20 µm2 for T. angustifolium to 134 µm2 for Tordylium maximum. Root CSA was correlated with the cortex plus rhizodermis CSA (r = 0.99, P < 0.001) and, to a lesser extent with the xylem CSA (r = 0.60, P < 0.05), but not its proportion (r = 0.15, P = 0.62). Interspecific variation in root CSA was not related to that in mean xylem vessel CSA (r = 0.40, P = 0.156), which was larger in Apiaceae than in other species.

Absolute xylem CSA, and xylem CSA as a proportion of root CSA, were markedly higher in perennials than in species with a short life span (F = 14.8, F = 9.67, respectively, P < 0.01) whereas root CSA and mean xylem vessel CSA were not significantly different between life spans (F = 1.9, F = 1.6, respectively, P = 0.3) (Fig. 2a–c). No particular trend along succession was found within the Asteraceae or Poaceae for all the anatomical traits studied (Table 2).

Figure 2.

Relationships between root anatomical characteristics (root cross-sectional area (CSA), xylem CSA, and proportion of xylem in CSA) and (a–c) tissue mass density of roots (TMDr); (d–f) relative growth rate (RGRmax); (g–i) height at maturity (Hmax), for 14 herbaceous species. Solid lines denote significant relationships. Each point represents a species: annuals (open symbols), perennials (closed symbols); species from early (○), intermediate (▿) and advanced (□) successional stages.

Interspecific variations in anatomical features contribute diversely to those in TMDr, RGRmax and Hmax

We tested whether interspecific variations in root anatomical features were related to those in TMDr, RGRmax and Hmax (Fig. 2). No correlation was found between TMDr and root CSA (r = 0.16, P = 0.58; Fig. 2a). TMDr was highly related to xylem CSA (r = 0.84, P < 0.001; Fig. 2b) and to xylem proportion in root CSA (r = 0.93, P < 0.001; Fig. 2c). RGRmax was not related to root CSA (r = 0.08, P = 0.78; Fig. 2d), but was negatively related to xylem CSA (r = –0.65, P < 0.05; Fig. 2e) and to xylem CSA as a proportion of root CSA (r = –0.87, P < 0.001; Fig. 2f). Hmax was positively correlated with root CSA (r = 0.66, P < 0.01; Fig. 2g), cortex plus rhizodermis CSA (r = 0.60, P < 0.05), absolute surface of xylem (r = 0.80, P < 0.001; Fig. 2h) and xylem CSA as a proportion of root CSA (r = 0.62, P < 0.05; Fig. 2i). No significant relationship was found between TMDr, RGRmax, Hmax and the mean xylem vessel CSA (r = –0.28, P = 0.34; r = 0.48, P = 0.08; r = 0.30, P = 0.29, respectively).

We further investigated the dependency of RGRmax and Hmax on TMDr and root CSA using multiple linear regressions. Given that TMDr and root CSA were mainly independent, their inclusion in multiple regressions enhanced the fraction of explained variance relative to separate regressions (Table 3). Multiple regression explained 88% of the variation in RGRmax (P < 0.001). The contribution of TMDr was higher than the contribution of root CSA, which is only marginally significant (P = 0.052). For Hmax, up to 67% of the variance was explained by variations in TMDr and root CSA (P < 0.001; Table 3). Both TMDr and root CSA contributed significantly, and to a similar extent, to variations in Hmax.

Table 3.  Results of multiple regressions between relative growth rate (RGRmax) or height at maturity (Hmax) as dependent variables, and tissue mass density of root (TMDr, log-transformed) and root cross-sectional area (CSA, log-transformed) as independent variables
  n r 2 Regression coefficients
TMDrRoot CSA
  1. Significance: ***, P < 0.001; **, P < 0.01; *, P < 0.05; a, 0.05 < P < 0.1.

RGRmax140.88***−0.31*** 0.06a
H max 140.67**91.08*87.09**

Discussion

Tissue mass density of roots (TMDr) varied significantly with inherent relative growth rate (RGRmax) and maximum plant height (Hmax) (Fig. 1), two important whole-plant traits commonly used in defining plant strategy (Grime, 1979; Chapin, 1980; Grime et al., 1988; Tilman, 1988; Westoby et al., 2002). Within woody species, no correlation was found between root tissue density and species growth rate (Wright & Westoby, 1999; Comas et al., 2002; Comas & Eissenstat, 2004). In contrast, within perennial grasses, TMDr was found to be associated with RGRmax but not with Hmax (Wahl & Ryser, 2000). In the present study, the presence of annual and perennial herbaceous species that exhibited contrasting TMDr might account for the strength of our correlations. The range of life spans studied thus enables us to detect a convergence between TMDr and plant functioning. At the above-ground level, a tight negative relationship was also reported between leaf TMD and RGRmax (Garnier, 1992; Poorter & Bergkotte, 1992; Ryser & Aeschlimann, 1999), stem TMD and RGRmax (Enquist et al., 1999), and stem TMD and Hmax (Enquist, 2003). These results provided additional evidence that the TMD of leaves and stems, but also of roots, are coupled and closely associated to whole-plant functioning. This is confirmed in our study, as TMDr was correlated with leaf TMD (r2 = 0.89, P < 0.001; data not shown) and with stem TMD, determined in natura (r2 = 0.84, P < 0.001; E.G., unpublished data).

How can interspecific differences in TMDr be explained? Xylem CSA, which is the most variable anatomical trait among our species, is a key determinant of TMDr (Fig. 2b). This result is in agreement with previous studies reporting that sclerenchymatic tissue area was associated with root TMD of 19 perennial grasses (Wahl & Ryser, 2000), leaf TMD of 14 grasses (Garnier & Laurent, 1994), and stem TMD of woody species (Castro-Diez et al., 1998; Preston et al., 2006). In our study, TMDr was positively related to xylem CSA and its proportion, but not to mean vessel CSA. These results suggested that differences in xylem CSA depend more on differences in the number of xylem vessels rather than in the size of vessels. High xylem CSA could result in a high tensile strength, which might protect roots against mechanical hazards and increase root longevity (Wahl & Ryser, 2000). It is thus not surprising that xylem CSA and its proportion were higher in perennial roots, compared with annual ones, and led to higher TMDr, as reported for 18 annual and perennial species from Argentina (Roumet et al., 2006).

RGRmax was negatively related to xylem CSA. In contrast, in perennial grasses, Wahl & Ryser (2000) found that RGRmax was associated to stele CSA rather than to xylem CSA. A reanalysis of Wahl & Ryser's (2000) data showed that this discrepancy might arise because xylem CSA is significantly correlated to stele CSA (r = 0.91, P < 0.001), but not to TMDr. The higher xylem CSA in slow-growing species suggested that, in these species, more biomass is allocated to structures involved in support rather than to those involved in growth. The strength of the correlation between RGRmax and xylem CSA is enhanced when xylem CSA is expressed as a proportion of root CSA (Fig. 2). Variations in root CSA accounted only marginally and positively for those in RGRmax (Table 3). Root CSA is linked physically to exchange surface area. As emphasized by Enquist (2002), natural selection has resulted in optimal fractal-like vascular networks, which minimize hydrodynamic resistance yet maximize organism resource use by maximizing the scaling of surfaces where resources are exchanged with the environment. The marginal positive contribution of root CSA to RGRmax (Table 3) indicated that root CSA contributed only partly to the maximization of exchange surfaces. Greater specific root length, smaller diameter, and more root tips per unit root length have been reported to increase the volume of soil explored per unit biomass invested in fine roots (Comas et al., 2002; Comas & Eissenstat, 2004).

In this study, taller plants showed large root CSA and a high proportion in xylem CSA (Table 3). The positive relationship found between Hmax and root CSA is in line with previous studies, in which plant stature scaled isometrically with root CSA (Wahl & Ryser, 2000; Niklas & Enquist, 2002) and with stem CSA (Niklas, 1995). These results suggested that increasing root CSA would enable stem enlargement and taller species. Height imposes some constraints both on anchorage and on water uptake and transport. Tall plants transmit forces to roots in tension; this might alter mechanical properties of roots, such as stiffness and strength, which could be diameter- and density-dependent (Ennos & Fitter, 1992; Fitter, 2002). Water flow involves radial conductivity into the root, which in woody and herbaceous species is negatively correlated with root diameter and with the width of the cortex (Rieger & Litvin, 1999); and axial conductance for the flux up to the shoot, which depends on xylem vessel width (Doussan et al., 1998). As differences in root CSA were explained mainly by variations in cortical plus rhizodermis CSA, enlargement of root CSA in taller species could increase radial resistance to water flow, as observed previously (Huang & Eissenstat, 2000; Nicotra et al., 2002). In contrast with Wahl & Ryser (2000), no relationship was found between Hmax and mean xylem vessel CSA, suggesting that axial water flow is not increased in taller species. A trade-off is, however, imposed on xylem vessel diameter in terms of hydraulic capacity and prevention of embolism (Alder et al., 1996; Steudle & Peterson, 1998; Steudle, 2000; McElrone et al., 2004; Sperry & Hacke, 2004), which might be of first importance in Mediterranean species exposed to seasonal drought (Hukin et al., 2005). We conclude that tall plants have roots dense enough to sustain growth in height and large enough in diameter to enable stem enlargement.

Conclusion

The present study provides evidence of close relationships between whole-plant functioning, root structure and root anatomy in herbaceous species. These relationships operate independently of species’ life spans and botanical families. Although TMDr was measured under conditions very different from RGRmax, and at a different period from Hmax, it was closely related to RGRmax and Hmax, suggesting that these traits are species-specific and are linked to the characteristics of the habitats from which the species originated. We have demonstrated that these relationships have a strong anatomical basis, as they are dependent on the proportion of different types of tissue in the root. Slow-growing species and tall plants exhibited dense roots that might increase root resistance and longevity. Their higher density is caused partly by the large fraction of the root CSA occupied by xylem. This suggests that root anatomy – especially root xylem CSA, the most variable anatomical trait – strongly constrains root structure and plant functioning. As found previously in leaves (Garnier et al., 2004; Kazakou et al., 2006), root tissue mass density appears to be a powerful functional marker of plant functioning. There is now an urgent need to test whether interspecific differences in TMDr and root anatomy are related to the main root functions such as hydraulic conductivity, nutrient and water uptake, and root longevity.

Acknowledgements

The first author was supported financially by the Université de Montpellier II. We would like to thank Mickael Sagne and Lucy Pujolas for their technical help, the staff of the CEFE experimental field for their help in setting up the common-garden experiment, and the staff of the ‘Plate-forme histocytologie et imagerie cellulaire végétale’ (CIRAD-AMIS, UMR 1098), where root anatomy analyses were performed. This work was supported by the French National Program PNBC ‘Geotraits’. This is a publication from the GDR 2574 ‘Utiliterres’ (CNRS, France).

Ancillary