A novel A-Ci curve (net CO2 assimilation rate of a leaf –An– as a function of its intercellular CO2 concentration –Ci) analysis method (Plant, Cell & Environment 27, 137–153, 2004) was used to estimate the CO2 transfer conductance (gi) and the maximal carboxylation (Vcmax) and electron transport (Jmax) potentials of ageing, non-senescing Pseudotsuga menziesii leaves in relation to their nitrogen (N) content and protein and pigment composition. Both gi and the stomatal conductance (gsc) of leaves were closely coupled to Vcmax, Jmax and An with all variables decreasing with increasing leaf age. Consequently, both Ci and Cc (chloroplastic CO2 concentration) remained largely conserved through successive growing seasons. The N content of leaves, as well as the amount of ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) and other sodium dodecyl sulfate-soluble proteins, increased during the first three growing seasons, then stabilized or decreased only slightly afterwards. Thus, the age-related photosynthetic nitrogen use efficiency (PNUE) decline of leaves was not a consequence of decreased allocation of N towards Rubisco and other proteins involved in bioenergetics and light harvesting. Rather, loss of photosynthetic capacity was the result of the decreased activation state of Rubisco and proportional down-regulation of electron transport towards the photosynthetic carbon reduction (PCR) and photorespiratory (PCO) cycles in response to a reduction of CO2 supply to the chloroplasts’ stroma. This study emphasizes the regulatory potential and homeostaticity of Cc– rather than photosynthetic metabolites or Ci– in relation to the commonly observed correlation between photosynthesis and gsc.
Few studies have investigated the physiological processes causing the age-related PNUE decline of evergreen leaves. Niinemets et al. (2004, 2005) recently suggested that there is a possible gradual decrease in fractional investment of leaf N in photosynthetic machinery and reallocation towards structural cell wall components. However, they did not measure the amount of N allocated to those components directly, but instead based their estimates on photosynthesis model parameters derived from gas exchange measurements. Still, Takashima, Hikosaka & Hirose (2004) showed that low PNUE evergreen Quercus species allocated appreciably more N to cell wall proteins and less N to Rubisco than deciduous species of the same genus. In both groups, the amount of N allocated to cell wall proteins correlated positively with the amount of LMA whereas the trend was reversed for Rubisco. Although Takashima et al. (2004) made their measurements on current-year foliage only, the fact that LMA is generally found to increase with leaf age in evergreen species (Escudero & Mediavilla 2003; Niinemets et al. 2004, 2005) may indicate a concomitant increase in N investment to structural compounds.
This being said, Warren & Adams (2000) measured the N and Rubisco content of Pinus pinaster needles up to 5 years of age and did not observe any significant change in N allocation to Rubisco among the various foliar age classes. Moreover, Wendler et al. (1995) found no degradation of Rubisco in older, non-senescing Eucalyptus globulus leaves as they remobilized part of their N content to flushing new leaves during spring. In annual and perennial deciduous plants, excess Rubisco can be down-regulated and serve as a N storage protein until the onset of leaf senescence (Eichelmann & Laisk 1999; Cheng & Fuchigami 2000). Indeed, several studies have indicated that in evergreen species Rubisco may be synthesized in excess of what is required to support the realized photosynthetic rates under optimal field conditions (Warren, Adams & Chen 2000; Warren, Dreyer & Adams 2003a). However, in these studies the Rubisco requirement of leaves was calculated from their Vcmax estimated under the assumption that the CO2 transfer conductance from the sub-stomatal cavities to the carboxylation sites in the chloroplasts (gi) is sufficiently large to be ignored. Yet, most studies concerned with evergreen species have indicated that gi is small enough in these plants to cause a significant drop in CO2 concentration between the sub-stomatal cavities and the carboxylation sites (Loreto et al. 1992; Hanba, Miyazawa & Terashima 1999; Manter & Kerrigan 2004), a condition that leads to the underestimation of Vcmax if not accounted for (Ethier & Livingston 2004).
The question arises, then, as to whether or not age-related changes in gi may contribute appreciably to the photosynthetic decline of ageing evergreen foliage. In senescing leaves of herbaceous annuals and deciduous perennials, gi has been shown to decline in parallel to photosynthesis (Loreto et al. 1994; Delfine et al. 1999; Grassi & Magnani 2005). Although new studies (Niinemets et al. 2005, 2006; Warren 2006) confirmed that gi also follows photosynthesis throughout its decline in ageing evergreen leaves, they disagreed as to whether or not gi contributes significantly to the loss of PNUE. For broadleaf evergreens, Niinemets et al. (2005, 2006) estimated that the photosynthetic limitation imposed by gi increases with leaf age, whereas for P. pinaster, Warren (2006) argued that the age-related PNUE decline is solely due to the deactivation of Rubisco.
Recently, Ethier & Livingston (2004) presented a novel A-Ci curve analysis method that accounts for gi through a non-rectangular hyperbola version of the photosynthesis model of Farquhar, von Caemmerer & Berry (1980). Here, we apply this new method (1) to assess age-related differences in photosynthetic capacity and gi in Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco), an evergreen conifer known to be photosynthetically limited by gi under natural conditions (Warren et al. 2003b), and, by combining the photosynthesis model results with leaf protein assays, (2) to determine if age-related changes in PNUE are attributable to diffusive (i.e. gsc and gi) or biochemical (e.g. Rubisco content and activity) causes.
MATERIALS AND METHODS
Site description and sampling
Measurements were made from late September to mid-October 2001 on twelve 52-year-old coastal Douglas-fir trees sampled (one tree per day) from an even-aged stand located on the east coast of Vancouver Island, British Columbia, Canada (49°52′N, 125°20′W). Full details about the study site are given in Humphreys et al. (2003). The trees were accessed via a canopy access tower or with static climbing ropes. Two to three adjacent branch sections approximately 1 m in length were collected from each tree, either from the upper (six trees sampled for sun shoots) or lower (six trees sampled for shade shoots) canopy layer. This sampling design was adopted to highlight differences in LMA and gi among even-aged shoots in comparison to that among shoots of different age. Branch sections were collected during late afternoon on the day preceding the photosynthesis measurements. Upon cutting, the open branch end was immediately plunged in a water-filled plastic bag then re-cut. The collected branch sections were then transported to ground level where they were transferred to a holding container, re-cut under water, then left on the forest floor until placed under fluorescent lamps inside an instrumentation hut for at least 60 min before starting the photosynthesis measurements. A total of five shoots per tree, each representing a single current- to 4-year-old branch internode, were measured successively from the intact branch sections over a period of up to 12 h. Preliminary experiments performed on the scaffold tower established that gas exchange of shoots was unaffected by branch excision and remained so after overnight storage (data not shown).
Gas exchange measurements and integrating sphere apparatus
Shoot gas exchange rates were measured using an open gas exchange system with independent control of leaf chamber CO2 and H2O concentrations and air temperature (LI-6400; Li-Cor, Lincoln, NE, USA). To eliminate shading among needles resulting from direct illumination, the instrument’s standard 7.50-cm-diameter clear cylindrical chamber designed for short needle conifers (model 6400-05; Li-Cor) was enclosed in a 15.24-cm-diameter barium sulfate-coated integrating sphere (Fig. 1a) and modified as to consist of only light-transmitting or light-reflecting surfaces. The PPFD incident on the needles (two-sided) was estimated from the readings of light sensors positioned on the sphere wall calibrated against angular PPFD measurements taken around the stem of representative sun and shade shoots (see Fig. 1a). These angular measurements showed that although the overall (hemispherically integrated) intensity of the upwelling light reflected from the lower hemisphere was 64% of that coming from above, when adding the fluxes coming from any two opposite directions, the angular two-sided PPFD distribution measured around the shoot axis inside the clear chamber remained within 5% of the median value. Thus, this integrating sphere configuration, while preserving the uniformity of the two-sided diffuse light field, retained some of the polar asymmetry of the shoot’s natural light environment.
Needle temperature was estimated using an energy balance for which the amount of shortwave radiation (400–2600 nm) absorbed by Douglas-fir needles in the integrating sphere was determined. To this end, the integrating sphere light spectrum and Douglas-fir needle absorptance over the 400–1100 nm waveband were measured with a spectroradiometer (LI-1800; Li-Cor) (Fig. 1b). A custom-made, matte black sensor head cover with open slits in which individual Douglas-fir needles were recessed then illuminated from above was used to measure needle transmittance. Foliar reflectance measurements were made according to Moran et al. (2000). Needle absorptance from 1100 to 2600 nm was approximated using the Douglas-fir needle reflectance curve of Woolley (1971) and general leaf transmittance versus reflectance relationships derived from Gates et al. (1965). Correspondingly, the integrating sphere light spectrum beyond 1100 nm was approximated according to the known spectral output of quartz-halogen lamps adjusted so the calculated area under the whole spectral curve matched the total amount of shortwave radiation measured with a thermopile solarimeter (CM5; Kipp & Zonen, Delft, Holland). Radiation absorption coefficients derived from the above spectral curves (see Fig. 1b) were used in the instrument’s energy balance algorithm to set the needle temperature to 22 ± 0.5 °C during measurements using the estimated median value for the two-sided PPFD distribution around the shoot axis (corrected for the light attenuation caused by the shoot portion enclosed in the clear chamber) as initial input variable.
Following acclimation of a branch section under fluorescent lamps (PPFD ∼ 400 µmol m−2 s−1), the integrating sphere assembly was clamped onto a branch internode, taking care to enclose a single needle age class inside the shoot chamber. Light intensity inside the integrating sphere was then gradually increased to saturating levels (estimated median two-sided PPFD ∼ 1200 µmol m−2 s−1) and the shoot chamber’s CO2 concentration and relative humidity eventually stabilized at 360 µmol mol−1 and above 70%, respectively. The shoot was left to acclimate to these conditions before its steady-state An was recorded. Measurements of An under saturating light were then repeated at various chamber CO2 concentrations to obtain an An versus Ci relationship. After recording the A-Ci curve end point, the CO2 concentration of the chamber’s incoming air stream was reduced to a constant 1100 µmol mol−1 and An measured at various light intensities (decreasing the estimated median two-sided PPFD in steps from ∼ 1600 µmol m−2 s−1 to full darkness) to obtain a light response curve for the shoot. Ambient CO2 leakage inside the shoot chamber while being operated under low CO2 was largely prevented by wrapping the chamber’s foam gasket junction with Teflon tape and by keeping the air flow rate high enough to maintain sufficient positive pressure within the chamber to be applied against the under-pressurized air volume of the integrating sphere. Following gas exchange measurements, the shoot was excised from the branch section and immediately placed in liquid N for storage. Gas exchange parameters were calculated on a projected area basis according to the equations of von Caemmerer & Farquhar (1981).
Leaf photosynthesis model and curve-fitting procedure
where Ac and Aj are the RuBP-saturated and RuBP-limited net CO2 assimilation rate, respectively, Vcmax is the maximal CO2 carboxylation rate, J is the photochemical electron transport rate under RuBP-limited conditions, Rd is the mitochondrial respiration in the light, Γ * is the chloroplastic CO2 photocompensation point, and Kc & Ko are Michaelis-Menten constants for RuBP carboxylation and oxygenation, respectively, and O is the oxygen concentration. Detailed derivations of Eqns 2 and 3, as well as a thorough evaluation of errors resulting from assuming an infinite gi when fitting the Farquhar et al. (1980) model equations to A-Ci curves, are given in Ethier & Livingston (2004). Estimates of gi, Rd, Vcmax and J were obtained from non-linear least-squares fits (Levenberg-Marquardt algorithm) of Eqns 2 and 3 to the initial RuBP-saturated and remaining RuBP-limited A-Ci curve portion, respectively, taking care to exclude the end point(s) representing TPU-limited photosynthesis (Harley & Sharkey 1991), if observed. Contrary to established practice whereby the equation describing Ac is fitted to the data collected for Ci values below 200–250 µbar, we estimated that the transition from RuBP-saturated to RuBP-limited photosynthesis generally occurred around 400 µbar (see Appendix) and therefore used that value as the Ci cut-off point for fitting Eqn 2. The continuation of RuBP-saturated photosynthesis at Ci values greater than 400 µbar has previously been observed at ambient O2 concentration and high irradiance in Douglas-fir and other woody plant species with low gi (Manter & Kerrigan 2004). Furthermore, we noted in many cases that the curvature of the A-Ci relationship increased significantly when Ci approached the CO2 compensation point (Γ) (e.g. Fig. 2c). As discussed by Ethier & Livingston (2004), such an increase in the curvature of A-Ci curves at low CO2 is likely an indicative of Rubisco deactivation (von Caemmerer & Edmondson 1986; Sage, Sharkey & Seemann 1990) and of decreased re-fixation of respiratory CO2 by the enzyme (Pinelli & Loreto 2003). These data were excluded from the curve fits to avoid a possible overestimation gi. As with conventional A-Ci curve-fitting methods, the non-rectangular hyperbola model requires that Γ *, Kc and Ko be known a priori to estimate the remaining model parameters. The values used herein (Γ * = 39.5 µbar; Kc = 234.6 µbar; Ko = 256.4 mbar) were derived by Ethier & Livingston (2004) (see their table 3) from the original data of von Caemmerer et al. (1994) and adjusted to 22 °C using the temperature responses published by Bernacchi et al. (2002). These are to our knowledge the only complete data sets from which the kinetic constants of Rubisco have been properly evaluated at Ccin vivo. With these Rubisco kinetic constants substituted in Eqns 2 and 3, it was then possible to estimate Vcmax and J concurrently by iterating the least-squares fits to the RuBP-saturated versus RuBP-limited A-Ci curve portions described earlier until they produced converging solutions for gi and Rd. A comparison of gi values fitted by this method with alternative estimates of gi derived solely from the end points of the RuBP-limited A-Ci curve portion is presented in the Appendix. As shown therein, the agreement between the two methods is excellent, showing that our choice of respective domain for the Ac and Aj functions (see recent discussion) is justified.
To evaluate the bias introduced by assuming gi to be infinite when estimating Vcmax, the A-Ci curves’ RuBP-saturated portion was refitted with the common rectangular hyperbola model equation
using the aforementioned ‘chloroplastic’ Rubisco kinetic constants.
Light response curves were fitted as a non-rectangular hyperbola model (von Caemmerer 2000) in which J is given as
where I is the incident irradiance (incident two-sided PPFD around the shoot axis), α is the needle absorptance for the 400–700 nm waveband (0.9; determined from curve 2 in Fig. 1b), β is the fraction of absorbed light that reaches photosystem II (assumed to be 0.5), Φ is the quantum yield of photochemical electron flow photosystem II, Jmax is the maximal photochemical electron transport rate, and Θ is the convexity (curvature factor) of the rectangular hyperbola. Conversion from the original Aj measurements to J equivalents was done by rearranging Eqn 3 as
then solving Eqn 6 using the gi and Rd values derived from the corresponding A-Ci curve fit. Data collected at I < 50 µmol m−2 s−1 were excluded from the least-squares fits to avoid overestimation of ΦPSII as a result of the ‘Kok effect’ (Kirschbaum & Farquhar 1987). Transition from RuBP-limited to RuBP-saturated or TPU-limited photosynthesis at high irradiance which may occur when doing light response curves under ambient or saturating (e.g. 2000 µmol mol−1) CO2 concentrations, respectively, was avoided in our case by keeping the chamber CO2 concentrations around 1000 µmol mol−1 (and Ci between 600 and 800 µbar, depending on the shoot) during the high irradiance measurements.
Determination of leaf N, protein and pigment content
Needles from shoot samples stored in liquid N were weighed frozen and subsampled for determination of their projected area (measured by placing the needles between glass plates over a leaf area meter: LI-3100; Li-Cor) to fresh weight ratio and LMA (oven-dried). The rest of the needles were broken into coarse fragments and subsampled for Rubisco and total proteins analysis (see further discussion) or freeze-dried then ground in a ball mill (3110-3A Wig-L-Bug; Bratt Technologies LLC, East Orange, NJ, USA) for measurement of total N with an elemental analyser (Flash EA 1112 Series; ThermoQuest, Rodano, Italy) or for extraction of chlorophylls and carotenoids using N,N-dimethylformamide. Pigment solutions were assayed spectrophotometrically (Genesys 10; GENEQ Inc., Montréal, Canada) using the equations of Wellburn (1994) (1–4 nm range).
Crude leaf preparations were obtained by first homogenizing 50 mg of frozen fresh leaf tissue in 400 µL of ice cold buffer containing 133 mM TRIS pH 8.0, 25% (v/v) glycerol, 34 mM DL-dithiothreitol, 2% (w/v) polyvinylpolypyrolidone (PVPP), to which 100 µL of 10% (w/v) sodium dodecyl sulfate (SDS) was then added before vortexing for 4 min at 0–4 °C followed by centrifugation at 15 000 g for 5 min (4 °C). After collecting the supernatant, the pellet was re-extracted three more times with the said buffer (without PVPP) diluted 4:1 with 10% SDS and the supernatants pooled for the assay. Preliminary analysis of the protein composition of individual supernatants [quantification of the large subunit of Rubisco (Lsu) band on SDS-polyacrylimide gels by Coomassie brilliant blue R-250 staining] established the necessity of repeating the extraction fourfold and of adding a detergent (SDS) to the extraction buffer to ensure a thorough extraction (data not shown). Following extraction of all leaf samples, five additional ‘combined’ leaf extracts were prepared (to be used specifically for the preparation of Rubisco standards – see further discussion) by taking an aliquot from each leaf extract and pooling these according to foliar age class then diluting 1:500 using carbonate buffer (50 mM Na2CO3, 50 mM NaHCO3, 0.1 mM MgCl2.6H2O, pH 9.4).
Rubisco from the leaf extracts diluted 1:1000 in carbonate buffer was quantified by ELISA as described in Warren et al. (2003b). The Rubisco standards used herein were made from purified wheat Rubisco stock diluted to a concentration range of 0.052 to 6.6 µg mL−1 with the aforementioned carbonate buffer. The standards were then made specific to the chemical environment of the leaf extracts and to the various foliar age classes of the samples by suspending them into the said combined leaf extracts (1:1 volumes). All 12 samples from each leaf age class were then assayed according to their corresponding Rubisco standards. This was necessary as previous sensitivity tests showed that the standard curve for the ELISA estimation was affected by the chemical composition of the protein solution (see also Metodiev & Demirevska-Kepova 1992).
Following from previous work (e.g. Warren et al. 2000; Niinemets et al. 2004, 2005), alternative estimates of Rubisco concentration were calculated from the quotient of Vcmax/kcat where kcat is the assumed catalytic turnover rate of Rubisco. The value used herein (2.76 mol CO2 mol Rubisco sites−1 s−1) was determined in vivo by von Caemmerer et al. (1994)– the study on which our estimates of in vivo Rubisco kinetic constants are based – and adjusted to 22 °C according to the temperature response of Makino, Mae & Ohira (1988).
The total protein content of the leaf extracts was determined from gel densitometry as described in Warren et al. (2003b). Proteins were first purified by precipitation according to the method of Wessel & Flugge (1984) and the resulting pellets re-suspended in 100 µL of 2% (w/v) SDS.
Stable carbon isotope composition of leaf soluble sugars from field samples
In addition to the branch sections used for the photosynthesis measurements, several adjacent smaller twigs, each representing 5 years of incremental growth, were also collected at the same time to estimate the stable carbon isotope composition (δ13C) of recently fixed carbon from current- to 4-year-old shoots at each sampling location. The samples (pooled by age class) were stored in liquid N immediately after collection. Leaf soluble sugars were subsequently extracted from freeze-dried ground subsamples following the method of Brugnoli et al. (1988). The δ13C value of the dry sugar extracts was measured with a continuous flow isotope ratio mass spectrometer (Integra; Europa Scientific, Crewe, UK) at the Stable Isotope Facility of UC Davies. Carbon isotope discrimination (Δ) was calculated as
where δ13Ca and δ13Cp are the isotope compositions of the ambient canopy air (determined as –10‰ during daytime; Whiticar, personal communication) and plant sugar extract, respectively, relative to the Pee Dee Belemnite (PDB) standard.
where Ca and Cs are the CO2 concentrations of the ambient air and leaf surface, respectively, ab and a are the fractionations due to diffusion through the boundary layer (2.9‰) and through stomata (4.4‰), respectively, ai is the combined fractionation due to the dissolution and diffusion of CO2 in water (1.8‰), and b is the net fractionation caused by Rubisco and phosphoenolpyruvate (PEP) carboxylation (30‰).
Photosynthetic capacity and CO2 diffusion
Gradual loss of photosynthetic capacity with age was observed in both sun and shade shoots, the decline being equally apparent in the CO2 and light response of the shoots (e.g. Fig. 2). In some cases, the loss of photosynthetic capacity was eventually large enough to keep photosynthesis Rubisco-limited even at up to four times the normal daytime canopy air space CO2 concentrations (e.g. Fig. 2c). Strong linear relationships between gi and Vcmax were observed in individual foliar age classes (Fig. 3a), although the overall relationship for all age classes appeared to be slightly curvilinear. The slope of the relationships did not differ significantly among age classes (P > 0.5, analysis of covariance), suggesting that the sensitivity of carboxylation capacity to gi– or vice versa – does not change considerably with age. Yet, the Vcmax /gi ratio increased significantly with age (Table 1), hence reflecting an important age-related diminution of the internal CO2 supply to the chloroplasts relative to their collective carboxylation capacity. Within each foliar age class, the Vcmax /gi ratio remained conserved between sun and shade shoots (P > 0.1, unpaired t-tests). Carboxylation capacity (Vcmax) was consistently matched by the potential for photochemical electron transport (Jmax) throughout the range of gi values fitted by non-rectangular hyperbola model and across all age classes (Fig. 3b). Similar to photosynthetic capacity, the stomatal conductance to CO2 (gsc) was correlated with gi in all foliar age classes (Fig. 3c) and the slope of the relationship did not change significantly among the age classes (P > 0.5, analysis of covariance). However, unlike Vcmax and Jmax, the magnitude of gsc relative to gi decreased with age, essentially cancelling out the positive age effect of Vcmax /gi on the CO2 drawdown from the sub-stomatal cavities to the carboxylation sites (Ci –Cc) (Fig. 4a).
Table 1. Effect of needle age on the Vcmax to gi ratio of Douglas-fir shoots
Vcmax/gi (µmol mol−1)
Values given are the mean (SE) of 12 shoots. The significance of differences among age classes (P) was assessed by one-way analysis of variance (repeated measures). Where differences were significant (P < 0.05), Tukey HSD test was used to explore pairwise comparisons among age classes. Means followed by different letters are significantly different.
1 year old
2 years old
3 years old
4 years old
Area-based estimates of gi were consistently higher in sun shoots (Fig. 4b), although the difference was significant only in 4-year-old shoots (P < 0.05, unpaired t-tests). When LMA was factored in, the gi difference between sun and shade shoots disappeared (Fig. 4b–d), showing that, within each foliar age class, the response of gi to the shoot’s light environment is determined by changes in leaf anatomy. Over time, however, the courses of LMA and gi were mismatched – LMA increased steadily through successive years while gi decreased in steps after the second and fourth growing seasons (Fig. 4b–d). Variations in LMA were closely associated with parallel changes in both needle thickness (r2 = 0.67, P < 0.0001), needle dry to fresh weight ratio (r2 = 0.57, P < 0.0001), or tissue density (r2 = 0.56, P < 0.0001), and the slope of the relationships were conserved in sun and shade shoots (P > 0.5, analysis of covariance) whether evaluated among foliar age classes or within even-aged cohorts (data not shown). Yet, within each age class and over a similar range of values observed in sun or in shade shoots through time, mass-based estimates of gi did not correlate with LMA or the other anatomical traits (P > 0.3, data not shown).
Strong linear relationships between An and gsc evaluated at 360 µmol mol−1 ambient CO2 concentration were observed in all foliar age classes (Fig. 5a). The differences among slopes were small but significant in some cases owing to the close coupling of An and gsc in each age class. Consistent with the age differences in carboxylation capacity versus internal CO2 supply inferred from the A-Ci curve analyses, the slopes of the An versus gsc relationships scaled according to the ranking of the Vcmax /gi ratios fitted by the non-rectangular hyperbola model. Since gi was closely coordinated with gsc and An across the measured CO2 assimilation range, both Ci and Cc remained largely conserved among foliar age classes (Fig. 5b) and the corresponding Δgas exchange values calculated from Eqn 8 agreed well with the Δ value of sugars extracted from neighbouring field samples (Fig. 6). In fact, the isotopic composition of the field samples indicated an even greater conservativeness of Ci and Cc among age classes (Fig. 6b). Considering the shift of microclimate going from the field to the integrating sphere, this constancy of Δ is remarkable, especially for lower canopy shoots which went from mostly full shade to sustained saturating diffuse irradiance.
N allocation, Rubisco and leaf pigments
N per unit leaf area increased significantly with age in both sun and shade shoots (Table 2). In both cases, the accumulation of N was brought about by increases in mass concentration (Table 2) and by the steady rise of LMA (Fig. 4c). The fraction of needle N allocated to Rubisco was generally higher in sun shoots; the difference being significant in current-year shoots only (P < 0.05, unpaired t-tests), but did not differ significantly among foliar age classes (Table 2). N allocation to other SDS-soluble proteins increased slightly with age and, although the differences were not significant, they were sufficient (assuming 50 mol thylakoid N per mol chlorophyll –Evans 1989) to account for the observed increases in N allocation to chlorophyll-protein complexes (Table 2).
Table 2. Effect of needle age on PNUE of Douglas-fir shoots and the allocation of total nitrogen (Nmass, Narea) to Rubisco, other sodium dodecyl sulfate-soluble proteins, and chlorophylls (Chl)
Values given are the mean (SE) of six shoots. The significance of differences among age classes (P) was assessed as described in Table 1. Means followed by different letters are significantly different.
Despite their accumulating relatively high amounts of N, older shoots (≥ 2 years old) had significantly lower Vcmax per unit Rubisco than younger foliage and, as a result, had lower (up to 80% lower) PNUE values (Table 2). The loss of catalytic turnover rate of Rubisco (kcat) was considerable in 2- to 3-year-old shoots (25%), but particularly acute in 4-year-old shoots (68%), relative to the average kcat value of current- and 1-year-old shoots (2.81 ± 0.28 mol CO2 mol Rubisco sites−1 s−1) (Fig. 7a). Since the latter kcat value differs by less than 2% from our originally assumed value, the Rubisco concentrations determined from Vcmax and the kcat value of von Caemmerer et al. (1994) were essentially the same as those determined by ELISA for current- and 1-year-old shoots, but underestimated Rubisco in older shoots (Fig. 7b). By comparison, Rubisco concentrations calculated from Vcmax values estimated without considering gi (Eqn 4) underestimated the ELISA values in all foliar age classes (45–80% in older shoots –Fig. 7b).
There were significant correlations between kcat and the overall conductance to CO2 from the leaf surface to the stromal carboxylation sites (1.6/gs + 1/gi)−1 in ≥ 2-year-old shoots, whereas saturation of kcat was observed in younger foliage (mass-based relationships shown in Fig. 8). Both the slope and the regression coefficient of the relationships increased with needle age, suggesting an increasing degree of coupling between Rubisco activation and CO2 supply.
Loss of Rubisco activity in older foliage was matched by a proportional down-regulation of electron transport towards the PCR and photorespiratory (PCO) cycles (Fig. 9a) and by a corresponding increase in the carotenoid to chlorophyll ratio of needles (Fig. 9b). The correlations of kcat with J and the needle carotenoid to chlorophyll ratio were significant among foliar age classes for both sun and shade shoots (Fig. 9).
Leaf photosynthesis model curve fitting and estimation of gi
The non-rectangular hyperbola version of the model of Farquhar et al. (1980) contains an additional parameter (gi) used to quantify departures from the curvature of the original Michaelis-Menten kinetics rectangular hyperbola function valid at Cc (Ethier & Livingston 2004). Because (1) the addition of one more degree of freedom in the model introduces greater interaction among the fitted parameters and (2) the evaluation of the curvature of the RuBP-saturated and RuBP-limited A-Ci curve portions is usually done over a limited Ci range and is sensitive to experimental errors and/or significant changes in Rubisco activity, we recommend the use of a higher degree of measurement resolution (e.g. 25 µbar steps) over the domains of interest to increase the robustness of the corresponding least-squares fits. The A-Ci curves presented in this study were originally intended for use with the standard A-Ci curve-fitting procedure and, therefore, do not meet the said requirement. Thus, rather than treating the least-square fits of the RuBP-saturated and RuBP-limited A-Ci curve portions independently, we iterated the curve fits concurrently to obtain a unique solution for gi and thereby increase the number of measurements upon which the estimate is based. We assumed a constant gi throughout the measurement range; this has been partially confirmed (von Caemmerer & Evans 1991; Harley et al. 1992), but remains to be fully validated. Our parameter estimates depend on our choice of (1) Ci transition region between RuBP-saturated and RuBP-limited A-Ci curve portions (see Appendix), (2) stoichiometry of the equation describing RuBP-limited photosynthesis (Eqn 3), and (3) Rubisco kinetic constants. Given our methodology and assumptions, our estimates of gi for current-year shoots ranged from 0.059 to 0.144 mol m−2 s−1, which is on average 30% lower than the values obtained by Warren et al. (2003b) for the same study trees using a different method. We doubt that this is due to our choice of Rubisco kinetic constants as the values of Bernacchi et al. (2002) increased our gi estimates by only 7% while the commonly used in vitro values of Jordan & Ogren (1984) decreased them by 15%. Comparatively, changing the stoichiometry of Eqn 3 for the formulation which assumes that O2 is required as an additional electron acceptor to NADP+ via the Mehler reaction [i.e. Aj + Rd = J(Cc + Γ *) / (4.5Cc + 10.5Γ *) – see von Caemmerer 2000] increased our gi estimates by 13%, but resulted in greater initial disparity between the gi estimates of the RuBP-saturated versus RuBP-limited A-Ci curve portions to be iterated. The photocompensation point method of Warren et al. (2003b) is based on the intersection point of two A-Ci curves (high versus low irradiance), which occurs below Γ. As noted earlier, we found that, in many cases, the curvature of the An versus Ci relationship increases markedly in that region, which could influence the determination of the intercellular CO2 photocompensation point (Ci*) and Rd from which gi was estimated. Indeed, given the high relative sensitivity of the photocompensation point method to potential errors in Rd (i.e. potentially up to a 1:1 correspondence), the relative difference between our gi estimates and those of Warren et al. (2003b) is easily explained by the larger Rd estimates derived in the latter study (data not shown). Furthermore, Niinemets et al. (2005, 2006) recently established that the novel A-Ci curve-fitting method of Ethier & Livingston (2004) produced essentially the same estimates of gi as the chlorophyll fluorescence method.
Factors contributing to the PNUE decline of ageing Douglas-fir foliage
In Douglas-fir, the age-related PNUE decline of leaves begins after 2 years of growth and becomes severe after 4 years (Table 2). Because the proportion of leaf N allocated to Rubisco and other SDS-soluble proteins did not change appreciably throughout this period (or else increased in the case of chlorophyll-associated thylakoid proteins –Table 2), we reject the hypothesis that the PNUE decline resulted from a decrease in fractional investment of leaf N in photosynthetic machinery. Although we did not measure the amount of residual SDS-insoluble (presumably cell wall) proteins left in the final extraction pellet, we doubt that this protein fraction would have increased significantly with leaf age because Takashima et al. (2004) indicated that such an increase would be associated with a decrease in fractional investment of N to Rubisco and proteins involved in bioenergetics and light harvesting. According to an extensive foliage sampling at our study site in the year preceding our measurements, significant N resorption from senescing needles does not occur until the seventh growing season (data not shown). This surprisingly contrasts with our observation that the deactivation of Rubisco began during the third growing season (Fig. 7a) when the foliar N concentration was at its highest (Table 2) and continued thereafter. A similar pattern was found in P. pinaster (Warren 2006), hence confirming the role of Rubisco as a N storage protein in evergreen species (Warren et al. 2000, 2003a). Similarly, the age-related accumulation of chlorophyll-associated thylakoid proteins in spite of significant down-regulation of Rubisco and electron transport towards the PCR and PCO cycles (Fig. 9a) may reflect their role as potential N stores. Still, the concomitant increase of the carotenoid to chlorophyll ratio of needles (Fig. 9b) could instead indicate a shift in the function of protein-pigment complexes towards non-radiative dissipation of excess light energy via the xanthophyll cycle. For example, Ebbert et al. (2005) showed that overwintering sun- and shade-adapted Douglas-fir needles have the ability to down-regulate their photosynthetic capacity and up-regulate their capacity for xanthophyll cycle-dependent photoprotection. On the other hand, the work of Brooks, Hinckley & Sprugel (1994) and Brooks et al. (1996) on temperate conifers suggests that the age-related increase in N allocation to chlorophylls-protein complexes could be a consequence of acclimation to gradual shading. However, in these cases, shade acclimation was also indicated by a significant shift in pigment stoichiometry favoring chlorophyll b over chlorophyll a. By comparison, we found no significant age effect on the chlorophyll a/b ratio of Douglas-fir needles, whether collected from the upper (P = 0.2871) or lower (P = 0.6128) canopy layers (data not shown).
In this study, we found a threefold variation of Rubisco kcat values which affected Vcmax (Fig. 7) and to which Jmax and gi were correlated (Fig. 3a, b). In light of such results, we do not recommend the sole use of gas exchange parameters (i.e. Vcmax and Jmax) determined without considering the activation state of Rubisco (e.g. Niinemets et al. 2005, 2006), or Rubisco and gi (e.g. Ripullone et al. 2003; Niinemets et al. 2004), to explain age- or leaf life span-related differences in PNUE in terms of N allocation to Rubisco versus proteins involved in bioenergetics or light harvesting. Moreover, because gsc and gi both remained tightly coupled to the carboxylation efficiency of Rubisco over the years (Figs 3a, c & 8), we cannot exclude them as the primary cause of the age-related decline of PNUE (see further discussion).
Age-related decline of gi versus LMA and light acclimation
In Douglas-fir, LMA increases with irradiance during the first growing season (Aussenac 1973; Warren et al. 2003b), and then increases with needle age regardless of light in subsequent years (Ishii et al. 2002; this study) In both cases, the increase in LMA is modulated through changes in both needle thickness and tissue density, although in developing needles the changes mostly take place in the palisade mesophyll layer (Aussenac 1973), whereas in ageing needles, they are more likely to occur in the hypodermal layer and other lignified tissues (Apple et al. 2002). According to Niinemets et al. (2005, 2006) who discussed the possible effect of light acclimation versus leaf ageing on LMA and gi in Mediterranean broadleaf evergreens, the initial positive effect of light on LMA and gi (area basis) is a consequence of the greater chloroplast to total leaf surface area ratio of the ‘sun leaf type’ palisade tissue (see references given therein), whereas the subsequent negative effect of LMA on gi (mass or area basis) reflects the accumulation of cell wall compounds which would impede the diffusion of CO2 in the liquid phase. We agree that the positive effect of light on LMA and gi (area) in current-year foliage (Fig. 4b, c) is likely related to the development of a greater relative chloroplastic surface area. However, because the courses of LMA and gi were mismatched through time (Fig. 4b–d), we doubt that the overall negative correlation between LMA and gi (mass) we observed among foliar age classes reflects a direct causal link.
Coordination between Rubisco activity and CO2 conductances: a ‘chicken versus egg’ question
Over 20 years ago, Wong, Cowan & Farquhar (1979) were the first to propose that stomatal aperture is determined by the collective photosynthetic capacity of the chloroplasts such that gsc remains proportional to An, and Ci is kept nearly constant in the face of changing growth conditions that affect mesophyll photosynthesis directly (e.g. light, ambient CO2, N nutrition –Wong, Cowan & Farquhar 1985). Our results (Figs 5 & 6) confirm that An and gsc remain coordinated through successive growing seasons in evergreen leaves (Field & Mooney 1983; Yoshie 1986) which, according to Wong et al. (1979, 1985), would be the consequence of guard cells responding to the declining photosynthetic capacity of the mesophyll by reducing their aperture. Several photosynthetic metabolites have been proposed as possible mediators of the response, including ATP, NADPH and RuBP (Wong et al. 1979; Farquhar & Wong 1984), but many researchers have argued that Ci itself is the signal to which guard cells respond directly (Mott 1988; Lawson et al. 2002). If so, Rubisco deactivation would come first in the chain of events that leads to stomatal closure in ageing Douglas-fir leaves, the homeostatic response being mediated by a transient rise in Ci. Recently, however, von Caemmerer et al. (2004) challenged the theory of Wong et al. (1979) by showing that transgenic plants with reduced amounts of Rubisco and decreased photosynthetic capacity and gi (Evans et al. 1994) retain normal, wild-type guard cell characteristics including the rate of stomatal opening, steady-state gsc and gsc sensitivity to ambient CO2, but do not appear to perceive changes in Ci. Such results are difficult to reconcile with the view that gsc responds to changes in mesophyll photosynthetic capacity by sensing fluctuations in Ci. If loss of photosynthetic capacity due to Rubisco deactivation did not cause the gsc decline in ageing Douglas-fir leaves, what did?
Lately, the coordination between leaf photosynthetic capacity and xylem hydraulic conductance has been established at multiple scales including large tree trunks (e.g. Hubbard, Bond & Ryan 1999), branches (e.g. Brodribb, Holbrook & Gutiérrez 2002), whole seedlings (Hubbard et al. 2001) and leaf lamina or short-needle conifer shoots (Brodribb et al. 2005). Considering the serial positioning of stomata in the flow path of water and CO2 through leaves and the necessary dynamic coordination of the vapour and liquid phase resistances to water transport in plants (reviewed in Meinzer 2002), it is tempting to propose that the age-related stomatal and photosynthetic decline of Douglas-fir shoots is originally caused by a loss of hydraulic conductivity in ageing xylem pathways. In Douglas-fir stems, the hydraulic conductivity of xylem decreases with the age of annual wood rings (Spicer & Gartner 2001); this trait has been associated with a loss of hydraulic connection to leaves in ageing annual rings of maple tree branches (Melcher, Zwieniecki & Holbrook 2003) and with the formation of ‘hydraulic bottlenecks’ that become shoot abscission zones on older, pedunculate oak branches (Rust & Roloff 2002). In analogy to this, the addition of successive growth rings on older Douglas-fir branch internodes to supply the distal growth of younger foliage is likely to be associated with a loss hydraulic conductivity in the ageing, recessed vascular traces connected to old needles. Our observation that the diameter of the vascular connection of needles (determined from leaf scars examined under a dissecting microscope) gradually decreases with age could be a consequence of such hydraulic bottlenecking. More detailed microscopy studies are needed to verify this hypothesis.
CO2 supply versus the chloroplastic CO2 operating point
If we take the conservativeness of the N allocation to Rubisco and associated soluble proteins in the face of a declining hydraulic-stomatal capacity as an indication of the limited post-ontogenic phenotypic plasticity of the ‘evergreen’ carbon fixation machinery of ageing Douglas-fir needles, it becomes easier to appreciate the importance of Rubisco down-regulation to prevent the exhaustion of CO2 in the chloroplasts and ensuing impairment of the cycling of photosynthetic metabolites. For example, according to our calculations, had the Rubisco from 2- to 3-year-old needles remained fully activated, the operating Cc would have fallen close to Γ * at the measured gsc and gi values. In 4-year-old shoots, Cc would have been completely exhausted if the Rubisco activation state had increased to only 50% of its potential. Judging from the limited range of Cc values observed both in the field and in the controlled environment of the integrating sphere (Figs 5b & 6), it would appear that the homeostatic constraints of Douglas-fir photosynthetic metabolism set the lower operational limit for Cc (at least in non-droughted conditions) above 100 µbar. Because CO2 itself is involved in the regulation of the Rubisco activation state (via the reversible carbamylation of a lysine residue within the Rubisco’s catalytic site prior to the binding of Mg2+–Andrews & Lorimer 1987), the Cc decrease resulting from a significant loss of conductance to CO2 may indeed provide a simple feedforward mechanism by which the activation of Rubisco is reduced in proportion to the diminishing supply of CO2 (see Fig. 8). Interestingly, in transgenic plants with reduced Rubisco activity due to decreased expression of Rubisco activase, the senescence-related degradation of Rubisco is delayed relative to chlorophyll and other soluble proteins (He et al. 1997). Presumably, the persistence of large amounts of Rubisco late into leaf development is a consequence of the reduced carbamylation of the Rubisco pool (He et al. 1997); a similar mechanism may be involved in the age-related accumulation of Rubisco in evergreen leaves.
We have shown that, in Douglas-fir, the age-related PNUE decline of non-senescing leaves is not a consequence of decreased allocation of N towards Rubisco and other proteins involved in bioenergetics and light harvesting. Rather, loss of photosynthetic capacity was the result of the decreased activation state of Rubisco and proportional down-regulation of electron transport towards the PCR and PCO cycles in response to a reduction of CO2 supply to the carboxylation sites. This study emphasizes the regulatory potential and homeostaticity of Cc– rather than photosynthetic metabolites or Ci– in relation to the previously observed coordination between photosynthesis and stomatal conductance (Wong et al. 1979, 1985). Our results are consistent with the recent proposal that the integration of guard cell and mesophyll physiological responses could be the result of co-evolution rather than direct mechanistic linkage (von Caemmerer et al. 2004).
This research was supported by grants from the Natural Sciences and Engineering Research Council of Canada, the Canada Foundation for Innovation and the BC Knowledge Development Fund. We gratefully acknowledge Dr A.J. Keys for providing pure wheat Rubisco and Dr P. Millard for supplying polyclonal antibodies. We thank Dr J.A. Trofymow for the chemical analysis of litter fall at our study site, Dr M.J. Whiticar for the carbon isotope analysis of the canopy air, and Drs A.K. Mitchell and D.L. Spittlehouse for lending us instrumentation. John Christian, Ian Jacobs, Ivan Petrovic, Samantha Robbins and Lindsay White are warmly thanked for technical assistance. Elyn Humphries, Gordon Drewitt and Zoran Nesic also made many helpful contributions through their work at the site. Timberwest Forest Co. and Weyerhaeuser Canada Ltd generously provided access to and logistical support at the research site. Finally, we thank Dr Ü. Niinemets for his valuable and detailed comments on our original manuscript.
Determination of Ci cut-off point
Figure A1a is an idealized representation of the A-Ci curve shown in Fig. 2a, which serves to illustrate the theoretical principle upon which we based our determination of the Ci cut-off point where the transition from RuBP-saturated (Eqn 2) to RuBP-limited (Eqn 3) photosynthesis is expected to take place at 22 °C for Douglas-fir shoots exposed to a saturating diffuse irradiance. In this example, the transition takes place at around 480 µbar, between data points 6 and 7, after which the difference between the measured RuBP-limited An and that predicted from Eqn 2 increases with Ci. Hence, when moving the Ci cut-off point to higher values into the RuBP-limited portion of the A-Ci curve, the curvature of the remaining lower curve section is increased accordingly, and, consequently, the gi value estimated from Eqn 2 rises proportionally (see Ethier & Livingston 2004). Conversely, if the Ci cut-off point is lowered into the RuBP-saturated A-Ci curve portion, the curvature of the remaining higher curve section is reduced and the gi value estimated from Eqn 3 decreases proportionally. Table A1 gives the summary of model parameters fitted by Eqns 2 and 3 using the Ci cut-off points numbered 3 to 9 in Fig. A1a (see table footnote for details). As shown in Fig. A1b, the correct Ci cut-off point is found where the two gi values fitted by Eqns 2 and 3 converge. Thus, in non-ideal cases, provided that gi is sufficiently conserved over the respective domain of the RuBP-saturated and RuBP-limited A-Ci curve portions, the correct Ci cut-off point is expected to be found where the difference between the two gi values fitted by Eqns 2 and 3 is minimized. In this study, in the majority of cases, the difference between the two gi estimates was minimized when the Ci cut-off point was set around 400 µbar; corresponding model parameters derived as described were then used as starting values for the subsequent concurrent iteration of Eqns 2 and 3 (see Materials and Methods section for details).
Figure A1. (a) Idealized representation of the net CO2 assimilation rate (An) versus intercellular CO2 concentration (Ci) relationship shown in Fig. 2a; the RuBP-saturated (solid circles numbered 1 to 6) and RuBP-limited (open triangles numbered 7 to 12) An values were calculated from Eqns 2 and 3, respectively, with gi = 0.119 mol m−2 s−1, Rd = 1.12 µmol m−2 s−1, Vcmax = 55.3 µmol m−2 s−1 and J = 118.1 µmol m−2 s−1. (b) Difference between gi estimates obtained from least-squares regression fits of Eqn 2 versus Eqn 3 to the ideal A-Ci curve shown in panel (a) resulting from potential errors in the determination of the Ci cut-off point marking the transition between RuBP-saturated and RuBP-limited photosynthesis (see Table A1 for details).
Figure A2. Comparison between the CO2 transfer conductance (gi) of Douglas-fir shoots estimated from concurrent fitting of Eqns 2 & 3 to the RuBP-saturated and RuBP-limited A-Ci curve portions and that estimated from fitting Eqn 3 to the end points of the A-Ci curves’ RuBP-limited portion using alternative estimates of the rate of mitochondrial respiration in the light (Rd) based on Eqn A1 (see text for details).
Table A1. Sensitivity of non-rectangular hyperbola model parameters (true values indicated in bold characters) to errors in the determination of the intercellular CO2 concentration (Ci) marking the transition between RuBP-saturated and RuBP-limited photosynthesis for the ideal A-Ci curve shown in Fig. A1a
The first estimate of gi, as well as Vcmax and Rd values were obtained by fitting Eqn 2 to the data including and falling below the indicated Ci cut-off point, after which the second estimate of gi and the value for J were obtained by fitting Eqn 3 to the remainder of the data. The last two rows of the table show the effect of a ± 50% deviation of Rd from its true value on the estimates of gi and J obtained from Eqn 3.
See successive numbering of data points in Fig. A1a.
Alternative estimate of gi not requiring prior determination of a Ci cut-off point
Recently, Niinemets et al. (2006) showed that the overall effect of large errors in Rd (e.g. ± 50%) on gi estimation based on Eqn 3 is minor for leaves that have inherently low gi (e.g. < 0.2 mol m−2 s−1). This was also demonstrated earlier for a Douglas-fir shoot with an estimated gi of 0.119 mol m−2 s−1, showing less than 5% variation in gi for ± 50% deviation of Rd from its ‘true’ value (see Table A1). To obtain alternative estimates of gi requiring no a priori determination of a Ci cut-off point for the present study, we first derived a second set of Rd estimates for the shoots by fitting the original light response curve data (for I ≥ 50 µmol m−2 s−1) as
where Amax is the maximal, RuBP-limited gross CO2 assimilation rate and is the quantum yield of CO2 assimilation, then refitted Eqn 3 to the end points of the A-Ci curves’ RuBP-limited portion using the new estimates of Rd. Because the measured Aj from the light response curves approached CO2 saturation at low irradiance, we assumed that the Rd values fitted by Eqn A1 were little affected by differences in Cc. As predicted from the sensitivity analyses mentioned earlier, because the alternative estimates of Rd differed from our initial values by ± 20% at most (13% higher in average), we found excellent agreement between our initial gi estimates derived from concurrent fitting of Eqns 2 and 3 and the alternative gi estimates based on the end points of the RuBP-limited A-Ci curve portion (Fig. A2). Taken together, these results suggest that our choice of Ci cut-off point for the CO2 response curve of Douglas-fir shoots is realistic, as are our estimates of gi and other model parameters.