Leaf life span optimizes annual biomass production rather than plant photosynthetic capacity in an evergreen shrub

Authors


Author for correspondence:
T. Lamaze
Tel: +33 (0)5 61 55 69 36
Email: thierry.lamaze@cesbio.cnes.fr

Summary

  • Owing to nitrogen (N) translocation towards new leaves, the shedding of old leaves can increase the whole-plant carbon gain. It occurs when their photosynthetic nitrogen use efficiency (PNUE) declines below a given threshold.
  • Here, we investigated variations in net photosynthetic capacity (Amax), N resorption and PNUE in populations of Rhododendron ferrugineum presenting different mean leaf life spans (LLS).
  • Both populations had comparable annual leaf surface area production and Amax across leaf-age cohorts. Branch photosynthetic capacity was up to 95% higher in the population with the longer LLS mainly because of the high contribution of old leaves to the total leaf area. Despite lower N concentrations, old leaves maintained relatively high Amax and consequently PNUE that were higher than or similar to the values found in current-year leaves.
  • As the ratio of PNUE in old to PNUE in new leaves was always higher than the fraction of leaf N resorbed during leaf shedding, we concluded that leaf shedding did not improve plant photosynthetic capacity. We suggest that in R. ferrugineum, leaf shedding is mainly controlled by the leaf storage function and, therefore, that models aiming to explain LLS should not only consider the leaf carbon assimilation function, particularly in nutrient-poor habitats.

Introduction

Leaf longevity is involved in a suite of interrelated traits that characterize plant strategies. Extended leaf life span (LLS) has been hypothesized to be advantageous for nutrient conser-vation because it increases the mean residence time of the nutrients in the plants (Chabot & Hicks, 1982; Aerts, 1995; Garnier & Aronson, 1998). As a consequence, nutrient-poor habitats are often dominated by long LLS evergreen species (Chapin, 1980; Aerts, 1995; Jonasson, 1995b; Eckstein et al., 1999).

However, long LLS has also numerous drawbacks because it requires energetically costly structural reinforcement of the leaf (Escudero & Mediavilla, 2003) to make them less vulnerable to physical damage and herbivory (Westoby et al., 2000, 2002). As a consequence, plants with longer LLS generally have a lower specific leaf area (SLA), the light-capturing area per dry mass of leaf. This property reduces photosynthetic rate (Warren & Adams, 2004) especially by affecting leaf nitrogen (N) distribution and/or leaf CO2 conductance (Hikosaka, 2004, 2005). Light-saturated photosynthetic capacity (Amax) has also been shown to decrease with leaf age (Reich et al., 1991a; Kitajima et al., 1997; Mediavilla & Escudero, 2003). This decrease may be caused by protein dismantlement owing to sink–source relationships (Hikosaka, 2005). Furthermore, light interception is likely to decrease with leaf age as a result of overshading by new leaves. Thus, even if the photosynthetic capacity (Amax) does not decrease with leaf age, the photosynthetic rate of old leaves would be reduced by low irradiance (Oikawa et al., 2006). Accumulation of algae, fungi, debris and damage from herbivores have also been shown to reduce leaf photosynthesis with ageing (Westoby et al., 2000). Thus, it is essential to take into account the effect of leaf age on photosynthetic capacity in order to estimate the long-term carbon budget of the whole canopy (Kikuzawa, 1991; Kitajima et al., 1997). This is of particular importance for studies on evergreen species with long LLS because they may have a greater proportion of old leaves (Mediavilla & Escudero, 2003; Silla & Escudero, 2003; Reich et al., 2009).

Numerous studies have shown a strong linear relationship between the photosynthetic capacity and nitrogen content of the leaves (Evans, 1983, 1989; Reich et al., 1992; Karlsson, 1994a; Hikosaka, 2004, 2005; Pornon & Lamaze, 2007; Hidaka & Kitayama, 2009). The slope of the photosynthesis-nitrogen relationship differs significantly among species (Hikosaka, 2004; Warren & Adams, 2004), revealing differences in photosynthetic nitrogen use efficiency (PNUE). Variation in the slope is caused by several factors that have been well described (Hikosaka, 2004; Warren & Adams, 2004). Conversely, the variation of PNUE with leaf age has rarely been studied although it is of particular importance because it can determine the time when leaves are shed. In fact, it is widely believed that evolution has led plants to maximize their carbon gain by optimizing the distribution of nitrogen in the canopy (Franklin & Ågren, 2002; Anten, 2005; Anten & Poorter, 2009) and/or their leaf lifespan (Kikuzawa, 1991). From this perspective, it has been proposed that leaf shedding will occur when it increases the whole-plant carbon gain. This requires the carbon gain achieved with nitrogen retranslocated to the new leaf just before shedding to be higher than the carbon gain achieved by the old leaf (Franklin & Ågren, 2002). This condition is satisfied when PNUE in old senescent leaves, expressed as a fraction of that in new leaves, is lower than the fraction of leaf nitrogen resorbed during leaf shedding (RN) (Escudero & Mediavilla, 2003; Oikawa et al., 2008). Thus, it appears that both RN and PNUE can determine whether leaf shedding increases the whole-plant carbon gain. In several studies this condition has been proved to be satisfied. For example, Escudero & Mediavilla (2003) found a decline in photosynthetic capacity with leaf age for nine woody evergreen species without nitrogen decrease, revealing a decrease in PNUE with leaf age. For these species, leaf shedding actually occurred when it increased the whole-plant carbon gain.

One important issue concerning models based on whole-plant carbon gain maximization that aim to explain LLS or nitrogen distribution within the plant is to know whether they may be generalized to all plant species and habitats. Although the main function of leaves is carbon assimilation, several studies have shown that leaves, particularly long-lived leaves of evergreen species, play an important role in nutrient conservation and storage (Karlsson, 1994b; Lamaze et al., 2003; Warren & Adams, 2004; Pornon & Lamaze, 2007; Marty et al., 2009). In evergreen species, old foliage often supplies a large proportion of the nitrogen required to meet shoot growth (Karlsson, 1994b; Pasche et al., 2002; Lamaze et al., 2003; Silla & Escudero, 2003; Pornon & Lamaze, 2007; Marty et al., 2009). This storage function could be even greater than the carbon assimilation function for annual growth in nutrient-poor habitats. Karlsson (1994b) showed that, in a dwarf evergreen shrub, shading of non-reproductive branches had a marginal effect on the growth of the current-year shoots compared with the defoliation of the 1-yr-old leaf cohort, which removed nutrients susceptible to be translocated to growing shoots. Large amounts of nitrogen are translocated from old leaves to current-year leaves through leaf senescence and shedding but also via resorption from healthy/attached leaves and woody tissues, with kinetics depending on soil nitrogen availability and source–sink interactions (Marty et al., 2009). In nitrogen-poor habitats, the contribution of leaf shedding to nitrogen demand can be significant because root nitrogen uptake and shoot growth can be asynchronous (Pasche et al., 2002; Lamaze et al., 2003; Silla & Escudero, 2003; Marty et al., 2009). The major contribution of leaf shedding to shoot growth strongly suggests that, in these habitats, LLS could result from a trade-off between the necessity to meet nitrogen demand in growing shoots and to maximize carbon gain. Marty et al. (2009) found that LLS could vary significantly between populations of the evergreen alpine shrub Rhododendron ferrugineum according to soil nitrogen availability and the sink effect of shoot growth. They showed that on particularly low soil nitrogen, nitrogen resorption and leaf shedding were accelerated because of strong nitrogen demand for shoot growth. However, whether early leaf shedding increased whole-plant carbon gain and PNUE in addition to supplying new shoots with nutrients has not been investigated.

In the present study, we address this question by using the same model as Escudero & Mediavilla (2003). This model allowed Oikawa et al. (2008) to show that in Xanthium canadense leaf shedding could occur when it increased whole-plant carbon gain or not, depending on nitrogen availability. Moreover, the same authors showed that in X. canadense, nitrogen translocation towards new leaves was a primary factor that caused leaf shedding and that it was accelerated when nitrogen availability was low (Oikawa et al., 2005), probably to meet nitrogen demand in growing plant parts. Thus, it appears that carbon assimilation and nutrient storage are two leaf functions that can determine the time of leaf shedding in relation with external conditions. In agreement with Escudero and Mediavilla’s model, we propose that a high PNUE in old leaves (so that PNUEold leaves : PNUEnew leaves ratio is higher than RN) indicates that early leaf shedding sustains primarily shoot growth rather than optimizing plant photosynthetic capacity. To do that, we investigated changes in leaf photosynthetic capacity and PNUE with leaf age in both populations, assessed the effect of a difference in LLS on the plant photosynthetic capacity and verified whether the difference in LLS between the two populations results from differences of PNUE in both old and young leaves. We also discuss how the conflicting functions of leaves (carbon assimilation vs nutrient storage) can interfere and disturb the cost–benefit approach to leaf carbon economy, particularly in such nitrogen-poor habitats, where the nutrient storage function of leaves is crucial.

Materials and Methods

Site and species studied

The site where the study was conducted was previously described in Marty et al. (2009). Briefly, it is located in the central French Pyrenees in the vale of Estaragne (42°48′ N, 0°9′ E), oriented north-east/south-west (opening to the north) and extending over 3 km between 1850 and 2500 m above sea level (a.s.l.). The vegetation is characteristic of the subalpine Pyrenees belt. It is composed of a mosaic of meadow, shrubs and trees with long heathland/meadow ecotones on a granitic, amphibolitic and schistic geological substrate. Soils are acidic (pH = 4.7 ± 0.1 (SD); total nitrogen, 0.5% ± 0.044 (SD); bulk density, 0.65 ± 0.099 (SD)). Snow cover usually persists from late October until early June. Thus, the vegetation period usually extends from early June to the end of October.

Rhododendron ferrugineum L. (Ericaceae) is a 70–80 cm high evergreen shrub, widely distributed in the Alps and the Pyrenees between 1600 and 2200 m a.s.l. (Ozenda, 1985). The two R. ferrugineum populations studied: PA and PB, have significantly different LLS (17.9 months and 21.5 months in PA and PB, respectively; Marty et al., 2009) and were c. 500 m apart on the north-west facing slope of the mid section of the valley (between 2100 and 2200 m a.s.l.).

Gas-exchange measurements

Photosynthesis and dark respiration were measured for leaves of the three different age classes with a portable photosynthesis system (LCi; ADC BioScientific Ltd, Great Amwell, UK) between July and September in both 2006 and 2007. Measurements (331 in total) were conducted during clear sunny days between 09 : 00 and 16 : 00 h solar time under ambient CO2 partial pressure, air temperature and relative humidity. Photosynthetic activity was measured under saturating ambient irradiance (photosynthetic active radiation (PAR) > 1000 μmol m−2 s−1) and dark respiration (Rd) after we had covered the leaf chamber with an opaque material. As the aim was to evaluate the mean maximum carbon assimilation rate (Amax) of each leaf age class within each population and the photosynthetic capacity of shrubs in order to compare the two populations, we performed the measurements on a maximum number of shrubs in each population rather than on a maximum number of leaves per shrub. Therefore, measurements were performed on one leaf of each age class per branch of 10–20 shrubs per day. In order to compare photosynthetic capacity between the two populations, we only used values obtained during two successive days under similar atmospheric conditions, that is, saturating PAR and similar temperatures in the leaf chamber. Gas-exchange measurements were performed on shrubs with low branch density, thus experiencing low self-shading.

Leaf and branch photosynthetic capacities

During each month, Amax and Rd measurements were conducted on the same number of shrubs in PA and PB. In both populations, we averaged all the values collected monthly to assess the mean Amax and Rd of leaves of each age class. The potential photosynthetic capacity of each leaf age class was estimated by multiplying mean Amax by the total leaf area of the cohort. In the study site, shrubs generally have two to three leaf cohorts simultaneously on the same branch. We assumed that within-branch shading was unlikely to strongly reduce light interception with leaf age for several reasons. First, all leaf cohorts in this species are located in the upper 15 cm of the canopy and leaf surface area is low. Second, leaf angles for the young cohorts are steep, which increases light interception for old leaf cohorts. Third, photosynthetic active radiation in the site (2200 m a.s.l.) was very high during the sunny days (often higher than 2000 μmol m−2 s−1) and likely to reach C3 saturating irradiance for photosynthesis (PAR > 700 μmol m−2 s−1) even for the oldest and most shaded leaves. Finally, measurements were performed on individuals with low branch density where old leaves are not or only slightly shaded. Thus, we used Amax values as an index to compare daily net carbon assimilation potential between leaf cohorts.

An estimation of the maximum carboxylation rate at 25°C (Vm0) was given by a one-point method using only the measurement of Amax, the intracellular concentration of CO2 (Ci), and a value of dark respiration (Rd) (Pornon & Lamaze, 2007). Parameters and their Q10 functions for temperature adjustment were derived from Collatz et al. (1991) and Sellers et al. (1996). Branch photosynthetic capacity depends on three main parameters: leaf surface area produced annually, LLS and Amax of each leaf age class. The first two parameters affect the branch leaf area and the last determines the potential rate of carbon assimilation of the leaf surface.

To assess the number of attached leaves at each date we randomly selected 20 seed-sired individuals of similar size in each site. Three shoots per individual (60 shoots in total) were randomly tagged and leaves of different ages (L0, L1, L2: current-yr, 1-yr-old and 2-yr-old leaves, respectively) were counted on 1 June, 15 July, 15 August, 15 September and the 20 October during each of the 2005, 2006 and 2007 growing seasons. All L0 collected on 15 August (i.e. at the end of the growing season) of the three years were scanned and their area measured with image j 1.36b (National Institute of Health, Bethesda, MD, USA). Branch photosynthetic capacity was calculated as follows:

image(Eqn 1)

(Sx (t),are the leaf area of the x cohort at time t; Ax(t), mean Amax of the x cohort at time t; the beginning of the growing season (i.e. June) is denoted t = 0). Sx depends on the number and the area of leaves produced each year by a branch and on leaf shedding pattern. The Amax values introduced in Eqn 1 were mean values calculated with all the measurements collected monthly for each cohort. These parameters were measured for both populations from June to October. As gas exchange measurements were performed on a large number of branches (one branch per shrub per measurement day) representative of each individual, we assumed that branch photosynthetic capacity represented the photosynthetic capacity of a plant canopy unit.

Leaf nitrogen and PNUE

For the two sites, four harvests per year (1 June, 15 July, 10 August and 10 September) were carried out during the 2006 and 2007 growing seasons. At each date, 10 leaves of each cohort per individual were collected from 10 randomly selected individuals. All the leaves collected were scanned and their areas measured with image j 1.36b (National Institute of Health, Bethesda, MD, USA). They were then dried for 72 h at 60°C and weighed to calculate the leaf mass per area (LMA). Afterwards, all the leaves of the same cohort of each shrub were pooled and ground to a fine powder (< 1 μm) to be analysed with a carbon–nitrogen analy-ser (Model NA 1500; Carbo Erba, Milan, Italy).

At each leaf census date, Narea (g m−2) (the nitrogen content per unit leaf area) of each leaf age class was calculated as the product of the corresponding Nmass (g g−1 DW) (i.e. the nitrogen concentration per unit mass) and LMA (g DW m−2). For each leaf age class, PNUE was estimated by dividing mean Amax (μmol m−2 s−1), calculated with all the values obtained during 1 month, by the corresponding mean Narea (g m−2). In order to assess the effect of Rd on PNUE, we also calculated PNUE as the ratio Vm0 : Narea.

Statistics and calculations

Within each population, Amax and Rd of the three leaf age classes were compared at each date with one-way ANOVA followed by a Tukey HSD test. In order to know the sources of variation of Amax, we performed an analysis of variance with population, leaf age class and period as main effects. Calculations were performed with r (R development Core Team, 2006).

The fraction of nitrogen resorbed during leaf shedding (RN) was calculated with data obtained in Marty et al. (2009). At each time, RN(t) was calculated as the difference between the nitrogen resorption efficiency, REFF(t) (i.e. the proportion of nitrogen resorbed from full leaf expansion to death), and NR(t) (i.e. the proportion of nitrogen withdrawn from an attached/healthy leaf between the end of growth and a given time t).

image(Eqn 2)

Results

Leaf surface area

The leaf photosynthetic surface area produced annually on each branch was slightly higher in PB than in PA (e.g. 12% in either 2006 or 2007) mainly because of higher individual leaf area in the former (1.48 ± 0.13 and 1.71 ± 0.41 cm2 in PA and PB, respectively, Table 1). The L1 area was 45% higher in PB at the beginning of the vegetation period (June, Table 1) and more than twice as high in September and October. The L2 area was also much higher in PB than in PA, especially in July and August. Differences in LLS between the two populations largely explain the markedly higher total branch leaf area in PB through the vegetation period.

Table 1.   Mean photosynthetic surface area (cm2) for each leaf cohort (L0, L1 and L2; current year, 1-yr-old and 2-yr-old, respectively) during the vegetation period of 2006 and 2007 in Rhododendron ferrugineum populations PA and PB
 JuneJulyAugustSeptemberOctober
PAPBPAPBPAPBPAPBPAPB
  1. Values are mean ± SD.

L00.0 ± 0.00.0 ± 0.013.4 ± 0.414.9 ± 7.612.2 ± 0.414.6 ± 7.811.6 ± 1.314.2 ± 7.411.4 ± 1.113.9 ± 7.2
L112.9 ± 5.518.7 ± 3.611.3 ± 5.516.4 ± 5.010.0 ± 5.414.7 ± 5.25.6 ± 2.911.7 ± 3.74.5 ± 2.811.0 ± 3.3
L25.9 ± 0.4512.9 ± 4.12.7 ± 0.9710.9 ± 4.41.9 ± 1.318.2 ± 3.10.0 ± 0.02.1 ± 0.60.0 ± 0.01.7 ± 0.4
Total18.8 ± 6.031.6 ± 7.627.4 ± 6.942.2 ± 17.024.2 ± 6.337.7 ± 16.117.2 ± 1.628.0 ± 10.515.9 ± 1.726.6 ± 10.1

Leaf gas exchanges

Three-way ANOVA revealed significant population, leaf age class and period effects on Amax (Table 2). Two of the three interaction terms were also significant (Period × Leaf age class and Period × Population × Leaf age class). From mid-July to mid-August, in both populations, Amax was higher in L1 than in L0 and L2. After mid-August, the Amax of L0 became similar to L1 and was significantly higher than L2 until the end of September (Fig. 1). The Amax of L1 remained similar in the two populations throughout the vegetation period. The Amax of L2 decreased faster in PA than in PB, which contributed to the significant population effect revealed by the ANOVA. Indeed, the Amax of L2 was more than five times higher in PB than in PA at the end of August. Thereafter, no more L2 leaves remained attached in PA.

Table 2.   Analysis of variance of photosynthetic capacity (Amax; μmol m−2 s−1) in Rhododendron ferrugineum measured during the 2006 and 2007 vegetation periods (= 331)
Source of variationdfMSFP
Population1361.122.044.01 × 10−6
Leaf age class2645.639.415.67 × 10−16
Period2134.28.193.41 × 10−4
Period × Population213.10.800.45
Period × Leaf age class4140.18.551.47 × 10−6
Population × Leaf age class236.82.240.10
Period × Population × Leaf age class343.82.670.04
Residuals30916.4  
Figure 1.

 Photosynthetic capacity (Amax) and dark respiration (Rd) of the three leaf age classes during a growing season (average of 2006 and 2007 values) in Rhododendron ferrugineum populations PA (a,c) and PB (b,d). L0 (circles), L1 (squares), L2 (triangles) and current-year, 1-yr-old and 2-yr-old leaves, respectively; Leaf age class values (mean + SD, = 10–34) were compared at each date with an analysis of variance (ANOVA) followed by a Tukey HSD test. Values not sharing the same letters are significantly different (< 0.05).

In both populations, dark respiration (Rd) was similar in L1 and L2 throughout the vegetation period whereas Rd was much higher in young L0 (July, Fig. 1). The difference then decreased and was no longer significant in August and September.

Leaf cohort and branch photosynthetic capacities

Over the vegetation period, branch photosynthetic capacity was up to 95% higher in PB than in PA (Fig. 2). In July, the L1 compartment accounted for more than 50% of the branch photosynthetic capacity in both populations. In September, L1 represented only 32% of the branch photosynthetic capacity in PA vs 45% in PB. In PA, the L0 compartment became the most important potential source of carbon for the plant earlier (August, 49%) than in PB (September, 49%). In PA, compartment L2 made a marginal contribution to branch photosynthetic capacity (< 7% in July and 3% in August). By contrast, in PB, the photosynthetic capacity of compartment L2 was higher than that of compartment L0 in July (27% and 20%, respectively) and still represented 18% and 6% of the branch photosynthetic capacity in August and September, respectively.

Figure 2.

 Branch photosynthetic capacity during the growing period in Rhododendron ferrugineum populations PA and PB. Light grey, L0 (current-year leaves); dark grey, L1 (1-yr-old leaves); black, L2 (2-yr-old leaves).

Relationship between carbon assimilation and leaf nitrogen

Taking all the data together (years, periods and leaf age classes), we found a significant positive linear correlation between Narea and both Amax and Vm0 (Fig. 3). However, Narea explained a much higher proportion of the variation in Vm0 (r2 = 0.56) than in Amax (r2 = 0.15). No population or interaction effects were found for either relationship (ANCOVA; data not shown), indicating that PNUE values in individual leaves were not very different between the two populations.

Figure 3.

 Relationship between Narea and photosynthetic capacity (Amax) (a) or maximum carboxylation rate at 25°C (Vm0) (b) in Rhododendron ferrugineum populations PA (closed symbols) and PB (open symbols). Current-year leaves, circles; 1-yr-old leaves, triangles; 2-yr-old leaves, diamonds. Regression lines: (a) = 3.04x+3.76 (r2 = 0.15. = 0.02); (b) = −7.9 + 32.0x (r2 = 0.56; < 0.001).

There was also a strong negative correlation between Rd and Amax for L0 (Fig. 4). No relationship was found between Rd and Amax for older leaf cohorts (Fig. 4).

Figure 4.

 Relationships between leaf dark respiration (Rd) and photosynthetic capacity (Amax) in Rhododendron ferruginem. Current-year leaves, closed circles; 1-yr-old and 2-yr-old leaves, open circles.

Nitrogen resorption and PNUE

Despite lower Narea (Table 3), L1 presented values of Amax higher than or similar to L0 during the whole vegetation period (Fig. 2). Consequently, PNUE tended to be higher in L1 than in L0 (40% and 30% in PA and PB, respectively, Table 3). We found that Amax declined more rapidly with time than Narea in leaves older than 1 yr, so PNUE in L2 tended to be lower than in L1 (30% in PA and 15% in PB). By contrast, in both populations PNUE values were quite similar in L0 and L2 (Table 3).

Table 3.   Characteristics of the different leaf age classes in Rhododendron ferrugineum populations PA and PB
 Nmass (%)Narea (g m−2)Amax (μmol m−2 s−1)PNUE (μmol s−1 g−1 N)Rd (μmol m−2 s−1)
L0L1L2L0L1L2L0L1L2L0L1L2L0L1L2
  1. L0, L1, L2: current-year, 1-yr-old and 2-yr-old leaves, respectively. Nmass, leaf nitrogen concentration per unit mass; Narea, leaf nitrogen content per unit area; Amax, photosynthetic capacity; PNUE, photosynthetic nitrogen use efficiency; Rd, dark respiration.

  2. Values are mean ± SD. Values that do not share the same letters are significantly different (< 0.05 – ANOVA followed by Tukey HSD test).

PA1.72 ± 0.151.32 ± 0.101.02 ± 0.042.97 ± 0.262.45 ± 0.091.79 ± 0.1410.52 ± 2.48 a12.12 ± 3.24 a6.20 ± 3.54 b3.56 ± 0.88 a4.96 ± 1.27 a3.42 ± 1.93 a 4.08 ± 1.64 a 2.15 ± 0.43 b1.81 ± 0.37b
PB1.66 ± 0.251.34 ± 0.071.10 ± 0.092.72 ± 0.312.52 ± 0.231.98 ± 0.2512.26 ± 2.56 a14.69 ± 2.67 a9.69 ± 2.27 b4.50 ± 0.76 a5.88 ± 1.45 a4.98 ± 1.56 a3.64 ± 2.09a1.52 ± 0.50b1.51 ± 0.38b

Marty et al. (2009) showed that nitrogen resorption in attached/healthy leaves (NR) occurred throughout the entire leaf life in a linear manner for both populations, the kinetics (the slope of the NR line), however, being faster in PA than in PB (Fig. 5a,b). By contrast, nitrogen resorption efficiency (REFF, i.e. the proportion of nitrogen resorbed from full leaf expansion to death) was shown to be similar for both populations. As a consequence, the fraction of nitrogen resorbed during shedding (RN) decreased regularly in leaves older than 14 months (Fig. 5c), the decrease being more rapid in PA than in PB. Unlike species that resorb no or little leaf nitrogen until leaf shedding, RN was much lower than REFF and decreased with leaf age in R. ferrugineum. For example in PA, the shedding of one L2 in July provided twofold less nitrogen than the shedding of one L1 (Fig. 5d).

Figure 5.

 Changes in nitrogen resorption of both attached (NR, solid line) and dead leaves (REFF, dashed line) in populations A (a) and B (b) with leaf age (see Marty et al., 2009). (c) Changes in RN, that is, the fraction of nitrogen resorbed during leaf shedding, which is calculated as the difference between NR and REFF curves, with leaf age in Rhododendron ferrugineum populations PA and PB; (d) relative importance of the amount of nitrogen released by L1 (1-yr-old) shedding compared with L2 (2-yr-old) shedding during the growing season in populations A and B. Closed and open circles, (PNUE)old : PNUEmax in PA and PB, respectively; closed and open squares, PNUE[Vm0]old : PNUE[Vm0]new in PA and PB, respectively.

The ratio of PNUE (calculated either as Amax : Narea or as Vm0 : Narea) in leaves older than 13 months (PNUEold) to the maximal PNUE value (PNUEmax) was never lower than 0.45 and thus was always markedly higher than RN in both populations even for L2 (PNUEold : PNUEmax > RN; Fig. 5c).

Discussion

Nutrient-poor habitats are often dominated by evergreen species. These plants have an intrinsically low photosynthetic capacity because of anatomical and biochemical traits related to long LLS, such as small relative leaf nitrogen allocation to the photosynthetic apparatus (Poorter & Evans, 1998; Warren & Adams, 2000; Westoby et al., 2002; Takashima et al., 2004) or inefficient allocation because of an overinvestment in rubisco relative to other components of the photosynthetic machinery (Warren & Adams, 2004). Canopy photosynthesis depends on both leaf area and leaf photosynthetic characteristics. The present work reveals a large difference between the two populations of R. ferrugineum in branch photosynthetic capacity. Longer LLS in PB was the main factor explaining the markedly higher branch photosynthetic capacity in PB because the differences between the two populations in annual leaf surface area production and in Amax of the three leaf age classes were small. Higher LLS in PB resulted in significantly higher photosynthetic leaf surface area during the whole vegetation period (Table 1). During new shoot growth (July and August), old leaves (L1 and L2) provided 65% of the total leaf area in PB vs only 50% in PA. Because L1 and L2 maintained high values of Amax (Fig. 1), longer LLS strongly increased branch photosynthetic capacity and probably the carbon gain of the whole canopy. Thus, this result supports the hypothesis that being evergreen is advantageous in habitats with short vegetation period because by maintaining several leaf generations evergreen species increase their photosynthetic surface area and can fix carbon photosynthetically early in the vegetation period, before bud-break and new leaf expansion (Chabot & Hicks, 1982; Jonasson, 1995a).The presence of old foliage at the beginning of the vegetation period is particularly important for R. ferrugineum because L0 have lower Amax than L1 until the beginning of August (Fig. 1). Moreover, Pornon & Lamaze (2007) have shown that old foliage provides new leaves with photosynthetic products during shoot growth. The initial low Amax in young L0 was associated with a high respiration rate. Then, the subsequent increase in L0 Amax was associated with a corresponding decrease in the respiratory rates suggesting that growth respiration was the main factor limiting carbon acquisition in young leaves (Fig. 4).

As in many other studies (Evans, 1989; Reich et al., 1991b, 1992; Karlsson, 1994a), we found a positive linear relationship between Amax and Narea (Fig. 3) across leaf age classes. However, this relationship, although significant, was not very strong (r= 0.15, P = 0.02) possibly because measurements were performed outdoors where stomatal resistance can vary between measurements and affect Amax. This is supported by the fact that the relationship is found to be much stronger using the stoma-independent maximal capacity of rubisco for carboxylation (Vm0) rather than Amax. Moreover, it has been shown that a smaller fraction of leaf nitrogen is allocated to the photosynthetic machinery in evergreen species compared with deciduous species, which results in a weaker relationship between Amax and Narea (Hikosaka, 2004; Warren & Adams, 2004). As shown by the ANOVA (Table 2), leaf age class had a significant effect on Amax. As there was a positive relationship between Narea and Amax, a part of this effect was probably caused by the decrease in leaf nitrogen concentration (Narea and Nmass) with leaf age. However, it seems that leaf age class had an effect on Amax independently of leaf nitrogen concentration. Indeed for both populations, despite a lower Narea, L1 had a greater and L2 a similar Amax to young L0 (Fig. 1). The fact that in L0, high leaf nitrogen concentrations (Nmass and Narea; Table 3) were associated with a high respiratory rate (Rd) and relatively low photosynthetic capacity (Amax) suggests that a large part of cell nitrogen was invested in respiratory machinery (e.g. mitochondrial enzymes). The increase in Amax of young L0 was associated with a parallel decrease in respiration accompanied by some nitrogen resorption (Pornon & Lamaze, 2007; Marty et al., 2009). This suggests that some of the leaf nitrogen contained in the respiratory machinery was gradually translocated to other plant compartments. Thus, nitrogen translocation or withdrawal could result in an increased net assimilation of CO2 via a decrease in leaf respiration, and consequently to an increase in PNUE of L0 during the growing season.

Our results contrast with those of Escudero & Mediavilla (2003), who reported that leaf ageing had significant negative effects on photosynthetic rates per unit leaf area without a decline in nitrogen content per unit leaf area, resulting in a significant linear decrease in PNUE with leaf age. They found that the ratio of PNUE in the oldest leaves to PNUE in new leaves (PNUEold : PNUEnew) was close to but never significantly lower than RN (i.e. the fraction of nitrogen resorbed during leaf shedding). Therefore, they concluded that leaf shedding occurred when it increased whole-plant carbon gain (i.e. when PNUEold : PNUEnew became lower than RN). In the species they studied, RN was similar to REFF as there was no decline in leaf nitrogen content with leaf aging. Resorption of leaf nitrogen occurred only during the short period of time preceding leaf shedding. Thus, they found high RN for each species and the difference between PNUEold and PNUEnew did not have to be large as the model predicted an increase in the whole-plant carbon gain with leaf shedding. Conversely in R. ferrugineum, because the proportion of nitrogen resorbed during leaf shedding (RN) decreased linearly with leaf ageing and although Amax decreased slightly with time in leaves older than 1-yr, leaves conserved high PNUE until the end of their life and the ratio PNUEold : PNUEmax always remained higher than RN (Fig. 5c). Nevertheless, although self-shading was low, some shading probably occurred for the oldest leaf cohorts. Consequently, actual PNUE in old leaves might be lower than calculated and the ratio PNUEold : PNUEmax might be slightly lower. However, it is unlikely that self-shading could reduce this ratio under RN because the former was always two times higher than the latter in both populations (up to three times higher in PA, Fig. 5c). It is also unlikely that the difference in mean LLS between the two populations resulted from differences in self-shading. Thus, in contrast to Escudero & Mediavilla (2003), this result seems to indicate that leaf shedding in R. ferrugineum did not occur when it increased the whole-plant photosynthetic capacity. Another argument supporting this conclusion is that a non-negligible fraction of leaves fell after shoot growth (after August, Table 1). Thus, nitrogen resorbed from these leaves was transferred to woody tissues with nil PNUE (Marty et al., 2009). As proposed by Marty et al. (2009), earlier leaf shedding in PA could result from lower soil nitrogen availability during shoot growth so it was unable to meet nitrogen demand. During shoot growth (between mid-June and mid-July), the shedding of a 1-yr old leaf releases about two times more nitrogen than the shedding of a 2-yr old leaf (RN L1/RN L2 ≈ 2; Fig. 5d). Thus, higher shedding of young leaves (L1) in PA (i.e. lower LLS) could be a way to enable the plants in PA to meet nitrogen demand and thus optimize annual biomass production.

In conclusion, earlier leaf shedding in PA compared with PB significantly reduced branch photosynthetic capacity by reducing old leaf surface area with relatively high photosynthetic capacity. This result shows that old leaf cohorts play a major role in R. ferrugineum carbon budget and supports the hypothesis that in subalpine habitats, where the favourable period is relatively short and environmental conditions limit photosynthesis potential, long LLS considerably increases whole-plant carbon assimilation. Although, a large fraction of the nitrogen resorbed during leaf shedding contributed to the production of young leaf biomass (Marty et al., 2009), PNUE in young leaves was too low compared to that of old leaves to increase the whole-plant photosynthetic capacity. Thus, it seems that earlier leaf shedding (and consequently reduced LLS) in R. ferrugineum results from the need to meet nitrogen demand in growing shoots and occurs even though it strongly reduces whole-plant carbon assimilation capacity. In R. ferrugineum, the leaf-shedding pattern seems to be a major factor in enabling the plant to produce enough biomass to survive in nutrient-poor habitat rather than to increase the whole-plant carbon gain and to optimize PNUE. Our results show that models aiming to predict LLS should not consider the carbon assimilation function of leaves alone, particularly in nutrient-poor habitats. In these habitats, annual biomass production can be either carbon or nutrient limited, or both, depending on species, growth conditions or even branch reproductive status (Karlsson, 1994b; Jonasson, 1995a). Thus, leaf shedding patterns and leaf life span may be controlled by both leaf functions.

Acknowledgements

We thank Etienne Muller and Luc Lambs (Ecolab, Toulouse, France) for lending us the photosynthesis apparatus, Peter Winterton for correcting the English, Paola Chavez, Anne Dozières and Olivier Nave for technical support, and the three referees for their constructive comments on the manuscript. This work was financially supported by the European Interreg Program ‘FluxPyr’.

Ancillary