Nitrogen absorption by roots as a cause of interspecific variations in leaf nitrogen concentration and photosynthetic capacity


  • Y. OSONE,

    Corresponding author
    1. Nikko Botanical Garden, Graduate School of Science, University of Tokyo, Nikko, Tochigi 321-1435, Japan
      †Author to whom correspondence should be addressed. Present address: Department of Plant Ecology, Forestry and Forest Products Research Institute, Matsunosato 1, Tsukuba, Ibaraki 305-8687, Japan. E-mail:
    Search for more papers by this author

    1. Nikko Botanical Garden, Graduate School of Science, University of Tokyo, Nikko, Tochigi 321-1435, Japan
    Search for more papers by this author

†Author to whom correspondence should be addressed. Present address: Department of Plant Ecology, Forestry and Forest Products Research Institute, Matsunosato 1, Tsukuba, Ibaraki 305-8687, Japan. E-mail:


  • 1Optimal biomass allocation models predicted that an increase in specific absorption rate (SAR: nitrogen absorption rate per unit root dry weight) increases the optimal leaf N concentration (LNC) which maximizes whole-plant growth rates. From this prediction, we hypothesized that inherent differences in the N absorption ability of roots, which is represented by differences in SAR, causes interspecific differences in LNC and photosynthetic capacity (Pmax).
  • 2Four deciduous tree species and three herb species were grown under controlled conditions to test this hypothesis. Despite growing under the same soil N conditions, the SAR of these species differed more than sixfold, with the deciduous trees having a smaller SAR than the herbs. Consistent with the hypothesis, there were strong positive correlations between SAR, LNC and Pmax. Leaf dry mass per area and the LNC–Pmax relationship, factors correlated with leaf life span, differed little among the species, suggesting a small effect of differences in these properties on the variation in LNC.
  • 3Simulations were performed using an optimal biomass allocation model and parameter values determined from measurements of each species. These demonstrated that the observed variation in LNC could be explained largely by differences in the N absorption ability of the species.
  • 4Additionally, the causal relationship between N absorption ability and LNC, suggested by optimal biomass allocation models, was verified by manipulating the biomass allocation between roots and leaves of Morus bombycis using uniconazole, an inhibitor of gibberellin synthesis.
  • 5These results indicate a close functional link between N absorption ability and LNC, which would account for large variations in LNC and Pmax across species.


Leaf nitrogen concentration (LNC) differs widely among plant species (Field & Mooney 1986; Evans 1989; Reich et al. 1991, 1999; Reich, Walters & Ellsworth 1992; Aerts & Chapin 2000). For example, herbs are generally high in LNC, ranging from 20–60 mg g−1; deciduous and evergreen trees are lower in LNC, with ranges 15–40 and 7–30 mg g−1, respectively (Reich, Walters & Ellsworth 1997). As a large proportion of leaf N is in the photosynthetic machinery, such as RuBP carboxylase (Evans 1983), differences in LNC are generally accompanied by differences in light-saturated photosynthetic rates (Pmax) (Field & Mooney 1986; Evans 1989; Reich et al. 1991, 1992, 1999). Therefore, differences in LNC, together with differences in photosynthetic N-use efficiency, contribute to interspecific variation in Pmax, Field & Mooney 1986; Evans 1989; Reich et al. 1991, 1992, 1999).

Although causes of the interspecific differences in photosynthetic N-use efficiency have been extensively studied (Hikosaka et al. 1998; Poorter & Evans 1998), causes of the differences in LNC remain unclear. It is possible that leaf life span, or physical and anatomical leaf properties associated with life span, are related to variations in LNC. Reich et al. (1991, 1992, 1999) identified a general trend for plants with low LNCs to have longer-lived leaves and larger leaf dry mass per area (LMA). Based on these observations, Reich et al. (1997, 1999) proposed several as yet untested explanations for variations in LNC. One possibility is that leaves with a long life span require a greater proportion of carbon-rich structures, such as thick cuticles and dense cell walls, than N-rich cytoplasm to build sturdy leaves, increasing the C : N ratio in such leaves compared with more ephemeral leaves. Alternatively, the dense structure of leaves with a long life span reduces the rate at which CO2 is transferred to the site of photosynthesis. This could limit the use of available N in photosynthesis (smaller photosynthetic N-use efficiency), and would select against the combination of high leaf N concentration and dense leaf structures.

In addition to these leaf properties, interspecific differences in stem or root properties could also cause differences in LNC. Osone & Tateno (2003) developed a model to investigate how a stem or root property affected the optimal LNC, that is, the LNC that maximized relative growth rate (RGR). Their model follows that of Hilbert (1990), in which internal N concentration is related to biomass allocation between roots and leaves such that an increase in root allocation at the expense of leaves increases the plant N concentration. The model predicts that both an increase in stem biomass fraction in the shoot biomass (analogous to the ‘stem cost’ of Givnish 1982) and an increase in N absorption rate per unit root biomass (specific absorption rate, SAR) increases optimal LNC. Consistent with this prediction, Osone & Tateno (2003) demonstrated a positive correlation between the stem fraction and the LNC and Pmax of temperate herbs, which suggests that interspecific differences in the stem fraction could be a source of the variation in the LNC and Pmax of temperate herbs.

Plant species sometimes express different SARs, even when grown with identical N availabilities (Poorter et al. 1991; Reich et al. 1998; Wright & Westoby 2000). These SARs represent inherent differences in the N absorption abilities of the roots of these species. According to optimal biomass allocation models, these differences in SAR would cause interspecific differences in LNC if each species regulated its root : leaf ratio and LNC optimally. Several observations are consistent with this prediction. Poorter et al. (1991) reported that among herbaceous species from various habitats, those with smaller N absorption abilities tend to have smaller LNCs. Similarly, there are positive correlations between N absorption ability and LNC among boreal deciduous and evergreen trees (Reich et al. 1998) and Australian woody species (Wright & Westoby 2000). However, these studies were correlative, and it is not clear that the correlations result from a causal relationship between LNC and root N absorption, as proposed by optimal biomass allocation models.

Here we tested, in three steps, the hypothesis that differences in N absorption ability cause interspecific differences in LNC, and consequently in Pmax. First, root traits, leaf traits and their relationships were analysed in three dicotyledonous herbs and four deciduous trees. The herbs had significantly larger root N absorption abilities than the deciduous trees. As herbs and deciduous trees have similar leaf life spans, using these plants could minimize the effect on LNC of differences in leaf life span, and highlight the effect of differences in N absorption ability. Second, simulations using the model of Osone & Tateno (2003) were performed to evaluate if each species achieved optimal LNCs. Third, the causal relationship between LNC and N absorption ability proposed by optimal biomass allocation models was verified by manipulating biomass allocation between roots and leaves of Morus bombycis, a deciduous tree, using an inhibitor of gibberellin synthesis.


Here the basic principle of the model is outlined. Equations are given in the Appendix. In the model, the plant consists of leaves that assimilate C as a function of LNC, stems that provide mechanical support for the leaves, and roots that absorb N. The biomass increment at time t is a product of the total leaf area and net assimilation rate (NAR). NAR is a rectangular hyperbolic function of the area-based leaf N concentration. The increment of plant N at time t is a product of root biomass and SAR, which can be interpreted as the soil N availability or the N absorption ability of the roots. The stem function, the mechanical support of the leaves, is represented by the fraction of the stem biomass in the shoot biomass. A larger stem fraction represents greater mechanical stability of the shoot. New biomass is allocated to each organ according to the allocation coefficients P1 (allocation to shoots per total biomass increment) and P2 (allocation to stems per biomass allocated to shoots). Provided the other parameters are constant, an increase in biomass allocation to the roots (smaller P1) increases the internal N concentration. This provides a trade-off between the two components of RGR: NAR and leaf area ratio (LAR); increases in LNC resulting from increased biomass allocation into roots raise NAR, but decrease LAR. As RGR is a product of NAR and LAR, it reaches maximum at a certain combination of NAR and LAR, or at a certain P1. The LNC at this point is defined as the optimum.

As SAR increases, maximum RGR and optimal LNC increase, and optimal root : leaf ratio decreases (Fig. 1). The reason why optimal LNC increases with increasing SAR is as follows (see also Hilbert 1990). When SAR is small, a unit increase in N concentration requires a larger biomass investment to roots (i.e. increasing LNC is more costly) than when SAR is large. Thus a large increase in LNC decreases leaf area, and consequently RGR, when SAR is small. Therefore RGR is maximized at a smaller LNC (the optimal LNC decreases) when SAR is small.

Figure 1.

Relationships between specific absorption rate of nitrogen and maximum RGR (a); optimal area-based leaf N concentration (b); and optimal root : leaf ratio (c) simulated by an optimal biomass allocation model. P2 represents the fraction of stem biomass in the shoot biomass. Revised from Osone & Tateno (2003).

Figure 1 suggests that plants with smaller SAR can achieve maximum RGR if their root : leaf ratio is regulated such that their LNCs are smaller than in plants with higher N absorption abilities. Thus, if plants with different N absorption abilities behave optimally, their LNCs should differ. Differences in the stem fraction (P2) also affect the maximum RGR, optimal root : leaf ratio and optimal LNC.

Materials and methods

plant growth

The deciduous trees used in the experiment were Prunus sargentii Rehd.; Zelkova serrata (Thunb.) Makino; Salix bakko Kimura; and Morus bombycis Koidz. These inhabit open, fertile sites and often populate secondary deciduous forests in Japan. The herb species used were Chenopodium album L., an annual herb that is a pioneer in secondary successions; Polygonum cuspidatum Sieb. et Zucc., a perennial common in early stages in both primary and secondary successions; and Trichosanthes cucumerioides (Ser.) Maxim, an annual vine that is also a pioneer in secondary successions. Seeds of Z. serrata, S. bakko and M. bombycis were collected in Nikko, Japan. As P. sargentii seeds have a short life span, this species was grown from seedlings that emerged in spring 2001. Seeds of C. album were collected from the experimental farm of the Nikko Botanical Garden. Polygonum cuspidatum seeds were collected at Mount Fuji, and T. cucumerioides seeds were collected in Saitama, Japan.

Morus bombycis and C. album, which were used to parameterize the model, were grown during June–August 2003. The other species, P. sargentii, Z. serrata, S. bakko, P. cuspidatum and T. cucumerioides, were grown during June–August 2001. Plants were grown in a glasshouse at the Nikko Botanical Garden, Nikko, Japan (37°N, 139°E). The conditions in the glasshouse during the experiment were: photosynthetic photon flux density 65–75% of ambient; mean temperature 20·1 °C; mean relative humidity 75%. Seeds were sown in sand in plastic pots.

Morus bombycis and C. album were grown under five different N treatments, controlled using nutrient solutions. Following Hirose & Kitajima (1986), the composition of the nutrients other than N in these solutions was 3 mm K2HPO4, 1 mm MgSO4·7H2O, 3 mm CaCl2, 25 µm H3BO3, 2 µm MnSO4·5H2O, 2 µm ZnSO4·7H2O, 0·5 µm CuSO4·5H2O, 0·5 µm Na2MoO4· 2H2O, 20 µm Fe-EDTA. Addition of NH4NO3 to the basal solution yielded five solutions with different N concentrations (0, 0·2, 2, 5 or 20 mm). The other species were grown with the 20 mm N solution. Solution pH was adjusted to 6·0 with 1 m HCl. Early in the experiment, 300 ml of the appropriate nutrient solutions were added to each pot every 2 days. Later, nutrient solutions were supplied once or twice a day depending on the plant growth rate, to avoid nutrient depletion and maintain the supply of N proportional to plant size. Every 2 days the pots were flushed with tap water to avoid salt accumulation.

Morus bombycis and C. album were harvested four times at intervals of about 15 days in July and August 2003. The other species were harvested three times at intervals of about 15 days in July and August 2001. At each harvest, four to eight plants were harvested depending on the N treatments. The plants were divided into leaves, stems and roots. Leaf area was determined immediately after sampling using a flatbed scanner and image analysis software for leaf area (lia32, freeware The roots were further divided into fine (<1 mm diameter) and axial (>1 mm diameter) roots. A small portion of the fine roots was randomly sampled and scanned on a flatbed scanner. The length of fresh roots was measured using image analysis software based on nih image (Kimura, Kikuchi & Yamasaki 1999; Kimura & Yamasaki 2001). The specific root length (SRL, root length per root dry weight) was calculated from the root length and dry weight. The dry weights of each plant part were determined after oven-drying at 80 °C for 3 days. The total N contents of each part were determined using an automatic N/C analyser (NC-80, Shimazu, Kyoto, Japan).

On the day before the second and third harvests, light-saturated photosynthetic rate was measured. It was not measured for M. bombycis and C. album plants grown with 0 and 0·2 mm N because of their small leaf sizes. The CO2 exchange rates of fully expanded, young, leaves were measured early in the morning using an infrared gas analyser (PP Systems, Hitchin, Herts, UK) at an ambient CO2 concentration of 350 µmol mol−1, leaf temperature of 25 °C, relative humidity of 70%, and saturating irradiance. The leaves for which the CO2 exchange rates were determined were marked, and the N content of those leaves was determined at the time of harvest.

treatment with an inhibitor of gibberellin synthesis

According to optimal biomass allocation models, plants with smaller SAR should have a smaller LNC than plants with larger SAR, to maximize their RGR (e.g. Hilbert 1990; Osone & Tateno 2003). In other words, if LNC is increased in plants with lower N absorption abilities to match those in plants with higher N absorption abilities by increasing biomass allocation to roots, the RGR of these plants falls because of the larger requirement of root allocation. This can be verified if biomass allocation to roots of a plant with low N absorption ability can be artificially increased. Reduced gibberellin biosynthesis increases biomass allocation to roots at the expense of leaves (Nagel, Konings & Lambers 2001). Uniconazole-P (Sumitomo, Osaka, Japan), an inhibitor of gibberellin synthesis, was used to manipulate the biomass allocate of a deciduous tree, M. bombycis, to test whether the relationship between LNC, Pmax, RGR and biomass allocation was as predicted by the models.

Morus bombycis plants were grown in sand at 20 mm N under the same conditions as described above. The plants were treated with uniconazole after reaching ≈20 cm in height. Plants were divided into two groups, one of which was sprayed once per week with 200 ml uniconazole-P diluted to 10 mg l−1 with tap water; while the other was sprayed once per week with 200 ml tap water. Replicates from both groups of plants were harvested three times. At each harvest, the leaf area, root length, dry weight and N content were measured for four or five plants per treatment, as described above. The maximum photosynthetic rate was also determined on the day before harvest.

calculations and statistical analysis

SAR and NAR were calculated for plant N concentration and dry weight of the second and third harvests, assuming that changes in leaf area and root dry weights were exponential between harvests (Hunt 1982). Other values were calculated using data from the third harvest when the total biomass of plants was similar (5–9 g). Regression analyses were used to examine relationships between variables. For the relationship between mass-based and area-based LNC and Pmax, differences in slopes and intercepts of regression lines were tested using ancova. Tukey's tests were used for testing interspecific differences among means, and t-tests were used to evaluate group means of deciduous trees and herbs, and means of uniconazole-treated and untreated M. bombycis. All statistical analyses were carried out with spss software.


root properties

SAR at 20 mm N varied sixfold among the species (Fig. 2a). The deciduous trees had smaller SARs than the herbs. Species with larger SAR generally had higher SRL (Fig. 2b). The only exception was T. cucumerioides, which had the highest SAR, but an SRL as small as the deciduous trees.

Figure 2.

Specific nitrogen absorption rate (a) and specific root length (b) of four deciduous trees and three herbs grown with 20 mm N. PS, Prunus sargentii; ZS, Zelkova serrata; SB, Salix bakko; MB, Morus bombycis; CA, Chenopodium album; PC, Polygonum cuspidatum; TC, Trichosanthes cucumerioides. Error bar = SE. Different letters represent significant differences (Tukey's P < 0·05).

leaf properties

The LMA of plants grown with 20 mm N varied 1·6-fold (Fig. 3). There was no relationship between plant life form (herbs vs deciduous trees) and ranking in LMA. Prunus sargentii, a deciduous tree, had the largest LMA; M. bombycis, also a deciduous tree, had the smallest. The mean LMAs of the deciduous trees and herbs were not significantly different from each other.

Figure 3.

Leaf mass per area of four deciduous trees and three herbs grown with 20 mm N. Abbreviations as in Fig. 2. Error bar = SE. Different letters represent significant difference (Tukey's P < 0·05).

Both on mass and area bases, Pmax of M. bombycis and C. album were strongly correlated with LNC, and the Pmax–LNC relationships were not statistically different between the two species (slope F = 0·45, P > 0·05, intercept F = 0·29, P > 0·05 for mass; slope F = 0·01, P > 0·05, intercept F = 3·07, P > 0·05 for area) (Fig. 4). The other deciduous trees and herbs had Pmax–LNC relationships similar to those of M. bombycis and C. album. The deciduous trees generally had lower LNC and Pmax than the herbs.

Figure 4.

Relationship between leaf nitrogen concentration and Pmax for deciduous trees (closed symbols) and herbs (open symbols). •, Morus bombycis; ▴, Prunus sargentii; ◆Zelkova serrata; ▪, Salix bakko; ○, Chenopodium album; ▵, Polygonum cuspidatum; ◆, Trichosanthes cucumerioides. For M. bombycis and C. album, data obtained from three N treatments (2, 5, 20 mm N) were pooled. For other species means were plotted. Leaf nitrogen concentration was measured on the same leaves as Pmax. Regression lines: Y = 9420X + 253 (R2 = 0·71, P < 0·001) for mass basis of M. bombycis; Y = 8387X + 262 (R2 = 0·78, P < 0·001) for mass basis of C. album; Y = 10·87X− 4·21 (R2 = 0·88, P < 0·001) for area basis of M. bombycis; Y = 8·60X+ 7·48 (R2 = 0·83, P < 0·001) for area basis of C. album.

relationship between sar and the lnc andpmax

Consistent with the model's qualitative prediction (Fig. 1), LNC at 20 mm N was positively correlated with SAR at 20 mm N species (Fig. 5a,b). Reflecting the close relationship between Pmax and LNC (Fig. 4), the Pmax at 20 mm N was also closely correlated with SAR (Fig. 5c,d).

Figure 5.

Relationships between specific absorption rate and mass-based leaf nitrogen concentration (a); area-based leaf N concentration (b); mass-based Pmax (c); area-based Pmax (d). Each data point is a species mean. Regression line: Y = 1·36X + 0·02 (R2 = 0·92, P < 0·001) for mass-based LNC; Y = 39·6X + 0·83 (R2 = 0·64, P < 0·05) for area-based LNC. Y = 14889X + 337 (R2 = 0·69, P < 0·05) for mass-based Pmax; Y = 417X + 12·7 (R2 = 0·74, P < 0·05) for area-based Pmax.


Simulations were performed using the model of Osone & Tateno (2003) to determine if these species realized optimal LNCs, that is, if the observed variation in LNC could be explained by differences in SAR. The parameter values used in the simulations are summarized in Table 1. Measured SAR (Fig. 2), LMA (Fig. 3) and stem fraction for each species were used as parameters. As the measurements were performed with plants of similar size (5–9 g), the stem fractions were similar between the deciduous trees and the herbs (Table 1). The equation for the LNC–NAR relationship was obtained by regression of the pooled data of all species (Fig. 6). The equations for the LNC–stem N concentration and LNC–root N concentration relationships were also obtained by regression of the pooled data (Fig. 7). Such regressions are statistically incorrect, as more data were available for M. bombycis and C. album, which were grown in five different N treatments, than for other species, which were grown in only one N treatment. However, it was impossible to obtain regression lines or curves individually for the species grown in one N treatment because intraspecific variations in LNC were small for these species.

Table 1.  Parameter values and units of the four deciduous trees and three herbs used for the model simulations
  1. Values of LMA, P2 and SAR were measured values of each species; other values were obtained from pooled data for all species.

SARSpecific absorption rategN g−1 day−1Prunus sargentii 0·0052
  Zelkova serrata 0·0083
  Salix bakko 0·0155
  Morus bombycis 0·0163
  Chenopodium album 0·0245
  Polygonum cuspidatum 0·0269
  Trichosanthes cucumerioides 0·0335
P2Fraction of stem biomass in shoot biomassP. sargentii 0·36
  Z. serrata 0·41
  S. bakko 0·39
  M. bombycis 0·39
  C. album 0·49
  P. cuspidaum 0·30
  T. cucumerioides 0·40
LMALeaf mass per areag m−2P. sargentii48·4
  Z. serrata38·4
  S. bakko30·9
  M. bombycis24·7
  C. album30·6
  P. cuspidaum36·7
  T. cucumerioides35·4
cSN1Constants for stem N concentration 14·7
cSN2  −0·493
cSN3   0·0136
cRN1Constants for root N concentration  0·598
cRN2   0·0029
AmaxMaximum net assimilation rateg m−2 day−1  7·19
KmConstant determining initial slope of netg m−2  0·32
assimilation rate   
cMinimum LNC for assimilationg m−2  0·30
Figure 6.

Net assimilation rate as a function of area-based leaf nitrogen concentration. Symbols as in Fig. 4. For Morus bombycis and Chenopodium album, data obtained from five N treatments were pooled. Regression analysis was performed for pooled data for all species. Regression curves: Y = 7·18(X − 0·3)/{0·32 + (X − 0·3)} (R2 = 0·61).

Figure 7.

Stem nitrogen concentration (a) and root N concentration (b) as functions of mass-based leaf N concentration. Symbols as in Fig. 4. Data for all species were pooled. Regression relationships: Y = 14·7X2 − 0·493X + 0·0136 (R2 = 0·86, P < 0·001) for stem–leaf N concentration; Y = 0·598X + 0029 (R2 = 0·73, P < 0·001) for root–leaf N concentration.

The optimal LNC, on both mass and area basis, approximated the measured values for all species (Fig. 8a,b). Optimal root : leaf ratios were also close to the measured values, and generally decreased with SAR (Fig. 8c). These results suggest that, in these species, LNC and root : leaf ratio were apparently regulated optimally. Another simulation was performed in which LMA and stem fraction were the means of those of all species (LMA = 35, stem fraction = 0·39). This simulation eliminates the effect of interspecific differences in LMA and stem fraction on the optimal LNC (and root : leaf ratio), and shows the fraction of the observed variation in LNC (and root : leaf ratio) that can be completely explained by the differences in SAR. The effects of differences in LMA and stem fraction on LNC, which are represented by the difference in LNCs predicted in this (Fig. 8, bold lines) and former simulations, were small. As a result, the predicted LNCs showed a similar increase to the measured LNCs with increasing SAR. This suggests that the observed variation in LNC is largely explained by the variation in SAR.

Figure 8.

Measured and simulated mass-based LNC (a); area-based LNC (b); root : leaf ratio (c) for the four deciduous trees and three herbs. •, measured; ▵, optimal value of each species simulated by the model. Parameter values of the simulation are given in Table 2. Lines represent simulated LNC when LMA and stem fraction were assumed to be the average of the seven species (LMA = 35, stem fraction = 0·39). Error bar = SD.

uniconazole treatments

Uniconazole-treated M. bombycis plants had larger root : leaf ratios and plant N concentrations (PNC) than the controls (Table 2). To determine what caused the increase in PNC, PNC was factored into two components. From the balanced activity hypothesis, the PNC of a plant under exponential growth is described as:

Table 2.  Plant nitrogen concentration and biomass data of untreated and uniconazole-treated Morus bombycis
 Measured PNC (mg g−1)Theoretical PNC (mg g−1)Mass-based LNC (mg g−1)Root : leaf ratioSAR (mgN g−1 day−1)NARm (g g−1 day−1)LMA (g m−2)LAR (m2 kg−1)RGR (mg g−1 day−1))
  1. Theoretical PNC was also calculated by the equation: PNC = (RWR SAR)/(LWR NARm). NARm, mass-based NAR. Means ± SE. Different letters represent significant difference (t-test, P < 0·05).

Control25·2 ± 0·6a25·233·3 ± 1·4a0·481 ± 0·061a6·7 ± 0·3a0·128 ± 0·0037a47·8 ± 1·8a10·9 ± 0·55a66·4 ± 2·0a
Uniconazol32·3 ± 0·9b30·546·0 ± 4·4b0·976 ± 0·152b3·9 ± 0·2b0·125 ± 0·0016a52·1 ± 4·2b8·25 ± 0·87b53·7 ± 2·8b
image( eqn 1 )

where RWR is the root biomass per total plant biomass; LWR is the leaf biomass per total plant biomass; and NARm is the net assimilation rate per leaf biomass (Hilbert 1990). RWR/LWR, which represents the root : leaf ratio, was increased by uniconazole (Table 2). The other component of the PNC, SAR/NARm, was decreased by uniconazole, but to a smaller degree than the increase in the root : leaf ratio. As a consequence, the PNC of the uniconazole-treated plants calculated by equation 1 was larger than that of the control, and was similar to the measured PNC. These results suggest that the increase in PNC in the uniconazole-treated plants was largely caused by the increased root : leaf ratio. Area-based LNC and Pmax were also increased by uniconazole, but maintained a similar Pmax–LNC relationship to that observed (Fig. 9). Despite larger LNC and Pmax, uniconazole-treated M. bombycis trees had smaller RGR than controls, because the increased root : leaf ratio decreased LAR (Table 2).

Figure 9.

Relationship between area-based Pmax and leaf nitrogen concentration in control and uniconazole-treated Morus bombycis. +, control (CT) and uniconazole-treated (UC) M. bombycis. Other symbols are as in Fig. 4 and represent species means.


relationship between lnc, pmaxand n absorption ability

Although the species studied had similar leaf phenology, considerable variations in LNC and Pmax, on both mass and area basis, were observed among the herbs and deciduous trees (Fig. 5). Differences in the LMA and LNC–Pmax relationship were small between the species (Figs 3 and 4). Thus differences in physical and anatomical leaf traits (Reich et al. 1997, 1999) cannot explain the variations in LNC and Pmax in the present study. On the other hand, SAR differed up to sixfold between species (Fig. 2a). Consistent with our hypothesis that differences in N absorption ability cause variations in LNC and Pmax, SAR was positively correlated with LNC and Pmax (Fig. 5). Optimal biomass allocation theory predicts that plants with lower SAR can achieve the maximum RGR if they regulate their root : leaf ratio such that their LNC is smaller (Fig. 1). Simulations showed that most of the species achieved the optimal LNC and root : leaf ratio (Fig. 8), suggesting that the same causal relationship as assumed in the models explains the observed correlation between the observed SAR and LNC.

This causal relationship was also tested by manipulating biomass allocation in M. bombycis. As a result of the increased root : leaf ratio, LNC and Pmax of the uniconazole-treated M. bombycis increased to the levels observed in T. cucumerioides, the herb with the greatest LNC of all species examined, although those of control M. bombycis trees were less than those of herb species (Fig. 9). Despite its larger Pmax, uniconazole-treated M. bombycis had a smaller RGR than control plants, because the increased root : leaf ratios greatly decreased LAR (Table 2). This suggests that, if plants with lower N absorption abilities increase their LNC to that of plants with higher N absorption abilities, their RGR would decrease due to the large investment in roots. This is consistent with the causal relationship assumed. It also suggests that LNC and Pmax of M. bombycis depend on the regulation of biomass allocation between roots and leaves, and are not necessarily fixed by specific physical or anatomical properties of the leaves. All these results support the idea that interspecific differences in N absorption ability could cause interspecific variations in LNC and Pmax.

While we conducted the experiments with a small set of species with similar leaf phenology from similar habitats (open and relatively fertile), the correlation between photosynthetic properties and N absorption ability found here could also be general across species of contrasting habitats, leaf life spans and functional groups. Poorter et al. (1991) reported a positive relationship between N absorption rates and mass-based LNC for 24 herbaceous species from habitats of different soil fertility. A similar correlation was found between N absorption rates and mass-based LNC (and Pmax in Reich et al. 1998) for nine boreal deciduous and evergreen trees (Reich et al. 1998) and 28 Australian deciduous and evergreen trees (Wright & Westoby 2000). However, the relationship between N absorption abilities and area-based LNC (and Pmax) is not straightforward: a positive but weaker correlation in Reich et al. (1998), and no correlation in Wright & Westoby (2000). This may be due to the negative correlation between LMA and N absorption ability (Reich et al. 1998; Wright & Westoby 2000). Larger LMA of the species with lower N absorption abilities would reduce the interspecific difference in area-based LNC and Pmax, and therefore the strength of the correlation between area-based LNC, Pmax and N absorption abilities. Whether a positive correlation occurs between N absorption ability and area-based properties depends on the variation in N absorption ability relative to that in LMA. Here, and in Reich et al. (1998), area-based properties and N absorption ability were closely associated across species, because variation in N absorption ability was greater than that in LMA.

differences in n absorption ability

As SRL was generally larger in species with larger N absorption ability (Fig. 2), interspecific differences in SAR can be partly explained by the differences in root morphology or anatomy. A lower SRL can indicate a greater root diameter or tissue density (Ryser 1996). As N absorption is proportional to root length, thick roots with a smaller length per unit dry weight absorb less N per unit dry weight than thin roots (Nye & Tinker 1977). A high root-tissue density is correlated with the development of secondary cell walls (Eissenstat & Achor 1999; Wahl & Ryser 2000). Secondary cell walls, rich in suberin and less permeable to water, reduce the potential nutrient absorption rate per unit root dry weight (Eissenstat 1997). Therefore the interspecific differences in SAR would be caused partly by either a greater root diameter or tissue density; which of the two traits is responsible for the variation in SRL is unknown. However, these traits probably extend root longevity (Eissenstat 1992; Ryser & Lambers 1995; Ryser 1996; Eissenstat et al. 2000). Smaller N absorption abilities may reflect, therefore, a strategy primarily to extend root longevity.

ecological implications

The correlation between SAR and Pmax prompts the question: does SAR also affect NAR (daily photosynthetic gain), and thereby RGR, if it affects instantaneous photosynthetic rates? Optimal biomass allocation models predict that an increase in SAR increases maximum RGR (Fig. 1) through increases in NAR (data not shown) and LWR (represented in Fig. 1 by decreased optimal root : leaf ratio with increasing SAR). Comparison of these variables between species requires growth analysis with completely controlled and identical conditions, but our data provide support for this prediction. NAR was generally larger (Fig. 6), and root : leaf ratio smaller, for species with larger SAR (Fig. 8). However, differences in NAR between the species were smaller than those in Pmax because of the diminishing increase in NAR with LNC (Fig. 6). While most studies that have measured RGR and its components across species show RGR to be strongly correlated to SLA, they also usually show some correlation between RGR and NAR, or RGR and LWR (Poorter & Van der Werf 1998). Furthermore, Shipley (2002) indicated that the strength of the correlation between NAR and RGR increases with increasing irradiance, and demonstrated that NAR was the component that most strongly correlated with RGR at a photosynthetic photon flux density of 500 µmol m2 s−1. Such evidence suggests that differences in SAR could contribute to interspecific differences in inherent RGR, through its effects on NAR and LWR.

Growing evidence suggests that species with longer life span, thinner structures, and higher tissue N concentration, respiration rates and resource uptake rates above ground often have comparable of traits below ground (Ryser 1996; Schläpfer & Ryser 1996; Reich et al. 1998; Craine et al. 2001, 2002). However, there have been relatively few attempts to explain these parallels between leaf and root traits. The functional relationships proposed here among some of these traits could explain the mechanisms of these correlations. The correlation between LMA and SRL could be a product of parallel evolution. Low-productivity environments select for organs with thick, dense structures and longer life span, both above and below ground (Grime 2001). These traits would directly affect resource-uptake rates. Thick, dense leaves with higher LMA inevitably have smaller photosynthetic N-use efficiency and thereby photosynthetic rates (Hikosaka et al. 1998; Poorter & Evans 1998). Similarly, lower SRL reduces the N absorption rate (Eissenstat 1997). Lower N absorption ability, in turn, reduces mass-based LNC (and also area-based LNC in certain cases) by the causal relationship presented in this study. The low photosynthetic N-use efficiency and LNC together reduce mass-based Pmax (and also area-based LNC) in the species with higher LMA and lower SRL.


We thank H. Nagashima, N. Osada, H. Taneda and K. Hikosaka for valuable comments and discussions. We also thank S. Nemoto, H. Takahashi and the staff of Nikko Botanical Garden for assistance in the experiments. An anonymous referee provided helpful comments on an earlier draft. This research was supported by a fellowship from the Japan Society for the Promotion of Science (JSPS) for Japanese Junior Scientists.


Equations used in the model simulations are shown. For detailed explanation of the model, see Osone & Tateno (2003).

The whole plant biomass (W) is:

image(  eqn A1 )

where WL, WS and WR are the leaf, stem and root biomass, respectively.

The leaf area (AL) is:

image(eqn A2 )

where LMA is the leaf dry mass per area.

The absolute growth rate is a product of the net assimilation rate (NAR) and leaf area:

image( eqn A3 )

Newly produced assimilates are allocated to roots and shoots according to an allocation coefficient, P1 (relative allocation of total assimilates to shoots). The assimilates allocated to shoots are further partitioned between leaves and stems. The stem–leaf relationship was defined by using a coefficient, P2 (stem biomass per total shoot biomass). Therefore the absolute growth rates of the leaves (dWL/dt), stems (dWS/dt) and roots (dWR/dt) are:

image( eqn A4 )
image(eqn A5 )
image( eqn A6 )

where 0 < P1 < 1 and 0 ≤ P2 < 1.

The rate of N uptake is proportional to the root biomass:

image( eqn A7 )

where N is the nitrogen content and SAR is the specific absorption rate of the root (net N uptake rate per unit of root biomass). The SAR can represent soil N availability and N absorption ability of the plant roots.

The stem (SNC) and root (RNC) N concentrations are defined as functions of leaf N concentration (LNC):

image( eqn A8 )
image( eqn A9 )

Where the c terms are coefficients, N is partitioned among the three organs so that these relationships are satisfied.

The net assimilation rate (NAR) is a rectangular hyperbolic function of the area-based leaf N concentration (LNCa):

image( eqn A10 )

where Amax is the maximum net assimilation rate; c is the minimum N concentration necessary to maintain the cell (including the N contained in nucleic acids and housekeeping enzymes); and Km is a constant that determines the initial slope of the curve. LNCa is given by:

image( eqn A11 )