Respiratory flexibility and efficiency are affected by simulated global change in Arctic plants


Author for correspondence:

Ari Kornfeld

Tel: +1 650 223 6928



  • Laboratory studies indicate that, in response to environmental conditions, plants modulate respiratory electron partitioning between the ‘energy-wasteful’ alternative pathway (AP) and the ‘energy-conserving’ cytochrome pathway (CP). Field data, however, are scarce. Here we investigate how 20-yr field manipulations simulating global change affected electron partitioning in Alaskan Arctic tundra species.
  • We sampled leaves from three dominant tundra species – Betula nana, Eriophorum vaginatum and Rubus chamaemorus – that had been strongly affected by manipulations of soil nutrients, light availability, and warming. We measured foliar dark respiration, in-vivo electron partitioning and alternative oxidase/cytochrome c oxidase concentrations in addition to leaf traits and mitochondrial ultrastructure.
  • Changes in leaf traits and ultrastructure were similar across species. Respiration at 20°C (R20) was reduced 15% in all three species grown at elevated temperature, suggesting thermal acclimation of respiration. In Betula, the species with the largest growth response to added nutrients, CP activity increased from 9.4 ± 0.8 to 16.6 ± 1.6 nmol O2 g−1 DM s−1 whereas AP activity was unchanged.
  • The ability of Betula to selectively increase CP activity in response to the environment may contribute to its overall ecological success by increasing respiratory energy efficiency, and thus retaining more carbon for growth.


Arctic and boreal peatlands contain up to one-third of terrestrial carbon stocks (Gorham, 1991; Turunen et al., 2002), much of it locked in the permafrost. As permafrost thaws, CO2 is released from the organic soils and mineral nutrients are released into the soil solution. Whether the Arctic tundra becomes a source or a sink of atmospheric CO2 depends in part upon the extent to which plant growth increases in response to changing temperature, nutrient availability and local CO2 concentration (Hobbie et al., 2002). Mitochondrial electron transport may be a physiological factor mediating these ecological effects, because the energy-wasteful alternative oxidase pathway (AP) bypasses energy-conserving proton pumps of the cytochrome c oxidase pathway (CP) thus resulting in less ATP being produced per CO2 released to the atmosphere by the TCA cycle. For example, reduced carbon use efficiency due to nutrient deficiency has recently been attributed to increased AP activity (Sieger et al., 2005).

Several hypotheses have been proposed for the upregulation of AP activity in response to laboratory-induced conditions. For example, first, the AP may help maintain carbon balance by consuming excess carbohydrates (Sieger et al., 2005). Secondly, leaves receiving energy in excess of their metabolic energy requirements may also engage the AP to safely remove energetic molecules that could produce membrane-damaging reactive oxygen species (Millenaar & Lambers, 2003; Zhang et al., 2011). A third hypothesis suggests that the plant may also be able to conserve N, when it is scarce, at the expense of more abundant energy resources by producing alternative oxidase (AOX), which is ten times smaller, on a molecular weight basis, than the CP protein complexes (Noguchi & Terashima, 2006). All three hypotheses invoke an imbalance between received energy and metabolic needs: if energy is in excess, AP is upregulated. If energy is limiting, AP should be downregulated, as was detected in shade species or in species grown under low illumination (Noguchi et al., 2001), whereas AOX was upregulated when chloroplasts received too much energy (Yoshida et al., 2007). Other abiotic factors associated with increased AP activity or capacity include phosphorus (P) deficiency (Gonzalez-Meler et al., 2001; Juszczuk et al., 2001) and low temperatures, which could impede energy balance by lowering metabolic rates (Fiorani et al., 2005; Armstrong et al., 2008). Importantly, these effects have not been investigated in plants growing under field conditions or long-term environmental treatments. Thus, considerable uncertainty remains as to what extent these studies, conducted under carefully constrained laboratory conditions, can be generalized to species growing under natural conditions and exposed to long-term resource limitations.

Because of the important role of the Arctic tundra in global change, several field manipulations have been maintained for the past 20 yr at the Long-Term Ecological Research field station at Toolik Lake, Alaska (Arctic LTER) in order to help better predict and understand the effects of nutrients, light and warming on Arctic tundra ecology (Shaver & Chapin, 1980, 1995; Bret-Harte et al., 2001). These experiments have found that fertilization substantially alters ecosystem productivity and community composition, but that the effects vary with light availability: in plots with ambient light, Betula nana (dwarf birch) abundance and canopy height increase to the extent that it ultimately overshadows the other species in the plots (Chapin et al., 1995; Bret-Harte et al., 2001). Experimental shading suppressed the growth of Betula, however, resulting in Rubus chamaemorus (cloudberry) dominating those plots. Community-level responses as well as leaf trait- and leaf-level photosynthesis effects have been reported (Chapin et al., 1995; Chapin & Shaver, 1996; Bret-Harte et al., 2001), but we are unaware of any investigations into the relationship between the species growth response and foliar respiratory metabolism in this ecosystem.

Here we describe results of the application of a field methodology (Kornfeld et al., 2012a) to investigate respiratory electron partitioning in three species that responded differently to experimental manipulations at the Arctic LTER site. We also measured leaf traits; nutrient and carbohydrate status; mitochondrial size and density; and relative alternative oxidase (AOX) and cytochrome c oxidase (COX) protein concentrations to determine to what extent measured changes in respiration and electron partitioning correlated with these attributes. These data were used to investigate two primary hypotheses: (1) AP activity in the long-term field nutrient and environmental manipulation would decrease in response to fertilization or to shade; (2) relative abundance of AOX protein would decrease in response to fertilization or to shade (cf. Gonzalez-Meler et al., 2001; Juszczuk et al., 2001; Noguchi & Terashima, 2006).

Materials and Methods

Site description

The field site is part of a continuing field manipulation of upland moist acidic tussock tundra at the Arctic LTER (68°38′N, 149°34′W, elevation 760 m). Four replicate blocks, established in 1989, are each divided into several 5 × 20 m2 treatment plots separated by 2-m walkways (Bret-Harte et al., 2001; Boelman et al., 2005). Fertilized plots receive 10 g m−2yr−1 NH4NO3 and 5 g m−2yr−1 P2O5 each spring immediately following snow melt. Glasshouses constructed over subplots, raised temperature by an average of 4°C (Boelman et al., 2005). Similar structures covered with shade cloth reduced photosynthetically active radiation by 50%. We sampled six treatments representing a factorial assignment of nutrient and environmental manipulations: control plots (ctl), nitrogen and phosphorus addition plots (NP), shade houses (SH), glasshouses (GH) and shade house/glasshouses with NP addition (SHNP, GHNP). Three species were selected for this study because they dominated the canopy in one or more of the treatment plots: Eriophorum vaginatum L., a sedge; Betula nana L., a deciduous shrub; and Rubus chamaemorus L., a perennial forb.

Sampling procedure for respiratory incubations

Total respiration and electron partitioning were determined using the 18O isotope discrimination method (Guy et al., 1989) as modified by Kornfeld et al. (2012a) for field studies. Shoots harvested from individual plots were placed on ice in a cooler and carried to a field laboratory a short distance away. Leaf tissue or, in the case of Betula, whole shoots were sliced into sections no larger than 1 cm × 7 cm and sealed in 12-ml septum-capped vials (Labco, High Wycombe, UK) in the dark. A half-pellet of KOH in a protective holder removed respired CO2. After 2 h, air was transferred by syringe into pre-evacuated 3.7-ml storage containers (Labco) for later analysis. Tissue was divided among six incubation vials in the FW proportions of 1 : 2 : 3 : 4 : 5 : 6, thus providing a six-fold range of O2 consumptions. Sample size was determined based on previously measured respiration rates, such that the tissue in the 6-portion vial would consume c. 50% of the O2 in the headspace in 2 h. Total sample size was c. 2.5 g FM for Betula and Rubus, and 2.1 g FM for Eriophorum. Each block was treated as a single replicate, resulting in a maximum of four replicates per treatment. Eriophorum, however, was not available in sufficient quantity for respiratory incubations in SHNP, nor in all but one GHNP plot. Previous studies have shown that cutting and transporting herbaceous or woody tissue as done here does not affect respiration (Turnbull et al., 2005; Searle et al., 2011), as confirmed by preliminary analysis using the species measured here (K. L. Griffin, unpublished data). The four treatment blocks were harvested sequentially over 8 d, with treatment order within each block selected randomly, in order to reduce systematic bias due to climatic variations during the harvest period. Tissue was generally harvested in mornings, but it should be noted that the harvests took place during a period of 24 h sunshine.

Discrimination ‘end-point’ values for the AP and CP were measured by soaking leaf sections in KCN or salicylhydroxamic acid (SHAM), respectively, buffered with 2 mM CaCl2 and 20 mM HEPES pH 7.2. Inhibitor concentrations – determined by titrating one inhibitor in the presence of the other and measuring O2 uptake using a liquid-phase Clark-type electrode (Digital Oxygen System Model 10, Rank Brothers, Cambridge, UK) – were 20 mM KCN and/or 30 mM SHAM for Betula and Rubus, and 100 mM KCN and/or 30 mM SHAM for Eriophorum (Supporting Information Fig. S5). Leaves from a fertilized plot were immersed for 1 h in buffer containing one, both or neither inhibitor, after which they were allowed to dry for 1 h before incubation. Measured discrimination values in the buffer-treated leaves did not differ from untreated leaves (Table S1). Residual respiration in doubly-inhibited tissue, based on in-vial respiration values, was 11% for Betula and Rubus, and 26% for Eriophorum. Only Betula endpoints, however, were considered reliable for computing τa: discrimination values for Rubus decreased upon KCN treatment, and a narrow endpoint range for Eriophorum combined with large residual respiration cast doubt on those endpoints (Table S1).

Discrimination, electron partitioning and respiration computations

Storage containers were shipped to Christchurch, New Zealand, and stored at laboratory temperature (20°C) for 10 months before measurement. Isotopic composition of the respiratory air samples was measured in a Delta V Plus isotope ratio mass spectrometer (Thermo Electron, Bremen, Germany) as described in Kornfeld et al. (2012b). Measurements were corrected for contamination using O2-free vials (‘blanks’) as described in Kornfeld et al. (2012a), but because of the long storage time at 20°C, significant curvature remained in the data even after correction. Discrimination values, D, were therefore computed using a quadratic regression (Kornfeld et al., 2012a).

Electron partitioning through the alternative pathway, τa, was determined by linear interpolation (Guy et al., 1989):

display math(Eqn 1)

(Ds, Da and Dc, discrimination values for the sample, alternative pathway, and cytochrome pathway, respectively).

In-vial respiration rates (R) were computed based on the proportion of O2 consumed, (f*), tissue mass (mass g), vial volume (Vvial, ml), tissue volume (Vplant, ml), CO2 trap volume (Vother, ml), incubation time (t, s) and incubation temperature (T, Kelvins) according to the equation:

display math(Eqn 2)

(P, atmospheric pressure (atm); Rgas, ideal gas constant (0.082 ml atm K−1 mmol−1)). Tissue volume was estimated based on FM divided by the density of representative samples, measured by water displacement. Incubation temperatures varied from 20 to 26°C across the 3-wk measurement period. No correlation was found between D and incubation temperature (data not shown). Respiration rates adjusted to 20°C, R20, were computed using Q10 values measured during the same experimental period or the following year, according to the formula (Atkin et al., 2005a):

display math(Eqn 3)

Respiration by each pathway was computed as

display math(Eqn 4(a))


display math(Eqn 4(b))

(va and vc, the activities of the alternative and cytochrome pathways, respectively).

Leaf tissue analyses

Leaf area was measured on representative samples by using WinRhizo (Regent Instruments, Quebec, CA, USA) to analyse images scanned on a flatbed scanner. Tissue was dried in an oven at 60°C for a minimum of 2 d, before measuring dry mass. The tissue was then sent to the Lamont-Doherty Earth Observatory, Palisades, NY, where the samples were ground, weighed and packaged for elemental analysis to determine (CHN) (2400 Series II, Perkin-Elmer, Boston, MA, USA). Remaining ground leaf samples were bulked by replicate block and sent to the North Carolina State University Environmental and Agricultural Testing Service (Raleigh, NC, USA) to determine total phosphorus concentration via wet digestion. Mitochondrial assessments were made on leaves collected in late July 2009 and then fixed at 5°C in buffered glutaraldehyde (2% w/v in 0.05 M potassium phosphate buffer), placed in sealed glass 20-ml vials, and transported in insulated containers from Alaska to Lamont-Doherty Earth Observatory. Samples were post-fixed and sectioned, and mitochondria were then quantified as described previously (Griffin et al., 2001; Wang et al., 2004).

A second set of samples were frozen to –80°C, lyophilized for at least 24 h, and subsequently shipped to the University of Canterbury, Christchurch, New Zealand. The samples were ground to a fine powder in a ball mill and then stored at −80°C until assayed for AOX, COX and for carbohydrates. Leaf carbohydrate content was assayed colorimetrically using phenol: sulfuric acid according to the method of Tissue & Wright (1995). Soluble sugars were extracted directly from powdered leaf tissue in methanol : chloroform : water. Starch in the remaining pellet was dissolved by hydrolysis to glucose in 35% perchloric acid before conducting the assay. Total nonstructural carbohydrates (TNC) was computed as the sum of starch and sugar content.

AOX and COX concentrations were determined semiquantitatively using a Western blot protocol. Proteins were processed as described by Laemmli (1970) with modification. Freeze-dried and powdered tissue was extracted in a solution containing 10% glycerol, 60 mM Tris pH 8, 2% sodium dodecyl sulfate (SDS), 0.51 mM EDTA and 0.05 mM tris(2-carboxyethyl)phosphine (Bond-Breaker TCEP solution, natural pH, Thermo Scientific, Rockford, IL, USA). The mixture was heated to 95°C for 10 min, centrifuged to remove coarse cellular debris, and then separated by electrophoresis in a 4–12% gradient polyacrylamide gel (Life Technologies, Carlsbad, CA, USA) at 200 V for 45–60 min. Proteins were transferred to nitrocellulose membranes using the iBLOT system (Life Technologies), and immunoblotted using the Snap i.d. system (Millipore, Billerica, MA, USA) with anti-AOX monoclonal antibodies (Elthon et al., 1989) at a dilution of 1 : 660 followed by anti-mouse horseradish peroxidase (HRP) conjugate. Results were visualized by chemiluminescence (ECL Advance, GE, UK) using a cryocooled digital camera (CHEMI GENIUS2, Syngene, Cambridge, UK). COX was probed with anti-COX antibodies (Agrisera, Vannas, Sweden) at a dilution of 1 : 50 000 followed by anti-rabbit HRP conjugate. Representative images of protein bands for both species and proteins are shown in Fig. 3. Bands were quantified using the GeneTools software package (Syngene).

Statistical analysis

Statistics were computed using R v1.14 (R Development Core Team, 2011). Treatment effects were initially tested for significance using three-way ANOVA with explanatory variables: Species (S), Nutrient (N : ctl, NP) and Environment (E: shade, open, or glasshouse). If interaction between N and E was detected, two-way ANOVA was performed, using Species and Treatment (ctl, NP, GH, GHNP, SH, SHNP), to analyse treatment effects. Because of missing data, ANOVA was sensitive to the order of variables. Variables were therefore analysed both as S, N, E and S, E, N; the larger P value of these two analyses is reported here. Species was put in the first position because interspecific differences take precedence based on biological considerations (Hector et al., 2010). To further mitigate the effects of unbalanced data on ANOVA, the data were also analysed using Akaike's Information Criteria for small samples (AICc), an order-independent measure of residual deviances. Linear models representing the 19 valid combinations of S, N, E and their interactions were ranked according to AIC weight, wi, which indicates the probability that any given model best explains the data (Burnham & Anderson, 2002; Venables & Ripley, 2002). Post-hoc tests were computed using Tukey's HSD comparison of all pairwise combinations, adjusted according to Westfall (Bretz et al., 2011). The pairwise comparisons were also used to generate the ‘letter groupings’ indicating statistically significant groups in the figures and tables.

Residuals variance was generally not homogenous across species. Statistical tests were therefore weighted by 1/vars, the within-species residuals variance (Venables & Ripley, 2002). The weighted residuals were tested for normality using the Shapiro-Wilk test (Royston, 1995). Residuals for SLA, R20, and AOX/COX were transformed to normality using the Geary-Hinkley transformation (Hayya et al., 1975).

Uncertainty is reported here as mean ± SE. If values did not differ by species, the data were pooled using ANOVA. Uncertainty in τa and the associated AP and CP rates, va and vc, include uncertainty from three measurements: the two endpoint discrimination values, Da and Dc, and the sample discrimination value, Ds in Eqn 1. The combined uncertainties in τa, va and vc were therefore determined by propagating the uncertainty (ISO/IEC, 1995). Assuming no correlation between the component variables, xi, the combined standard error, s, for a function f (x1, x2, …, xN) is computed as:

display math(Eqn 5)

(∂f/∂xi, partial derivative of f with respect to the ith component variable, xi; sex,I, standard error for xi). For τa, the resulting formula after simplification is:

display math(Eqn 6)

For the pathway activities, Eqn 4 (a,b), which are in the form of ab, the combined uncertainty is:

display math(Eqn 7)

In order to confirm that the component variables are independent, note that Da and Dc are invariant relative to the sample measurements. Respiration and discrimination on the other hand are computed from the same samples and so Ds and R could, potentially, be correlated. A graph of discrimination vs respiration across all field samples, however, confirms that the variables are not correlated (Fig. S1).


Leaf traits and mitochondrial characteristics

Leaf traits were measured to determine how they correlated with community-level effects as well as with foliar respiratory responses. DMC and SLA responded to the treatments similarly (no Species interaction; Table S2) for all three species Eriophorum vaginatum, Betula nana and Rubus chamaemorus. On the other hand, the effects of Nutrient differed by Environment for these two measures (> 0.001 and = 0.048, respectively; Table 1). Fertilization of uncovered plots lowered the DMC by 9 ± 2% (< 0.001) relative to controls. Shade lowered DMC by 6 ± 2% relative to controls (= 0.02) and these two effects added to a 17 ± 3% reduction in SHNP plots (< 0.01). Leaf DMC in the GH and GHNP plots, on the other hand did not differ from the controls or from each other (> 0.4). Although a similar, inverse, pattern can be seen in the response of SLA to treatment conditions (Table 1) only the increase of SLA in SHNP was significantly different from the unshaded treatments (< 0.03).

Table 1. Measured leaf traits for Betula nana, Eriophorum vaginatum and Rubus chamaemorus
 DMC (%)SLA (m2 kg−1 DM)N (mg g−1 DM)P (mg g−1 DM)N : P (mg N mg−1 P)Starch (mg g−1 DM)Sugars (mg g−1 DM)TNC (mg g−1 DM)
  1. Dry matter content (DMC), specific leaf area (SLA), starch, sugars, total nonstructural carbohydrates (TNC), nitrogen (N) and phosphorus content (P) and N : P ratio. Numbers are expressed as mean ± SE (n) for n blocks. Multiple samples from within each block were averaged before these computations to avoid pseudoreplication; some P assays were combined from several plots due to insufficient material. Treatments are ctl (control), shade house (SH) and glasshouse (GH) with or without added fertilizer (NP). Letters indicate statistically distinct groups at the < 0.05 significance level. Significant for all traits except N, P and N : P are based on combined data across Species, since no Species interaction was detected.

ctl 39 ± 2 (4)a13.3 ± 1.1 (4)a27.8 ± 1.4 (4)a3.0 ± 0.2 (4)a9.4 ± 0.6 (4)a60 ± 7 (4)a80 ± 8 (4)a139 ± 10 (4)a
NP 36.5 ± 0.8 (4)b,c14.4 ± 1.6 (4)a,b36.4 ± 1.1 (4)b6.1 ± 0.3 (4)b6.01 ± 0.19 (4)b45 ± 4 (4)a74 ± 12 (4)a119 ± 16 (4)ab
SH 37 ± 2 (4)c,d15.0 ± 0.8 (4)b,c31 ± 3 (3)a,b3.61 (1)a,c9.89 (1)a,c41 ± 2 (4)a85 ± 13 (4)a126 ± 14 (4)b
SHNP 33.6 ± 2 (4)b16.8 ± 0.8 (4)c35.6 ± 0.9 (3)b7.0 ± 0.4 (3)b5.1 ± 0.3 (3)b45 ± 4 (4)a73 ± 12 (4)a119 ± 13 (4)b
GH 39.1 ± 0.9 (4)c,a14.3 ± 0.8 (4)a,b26.2 ± 1.0 (4)a2.77 ± 0.20 (5)a9.5 ± 0.9 (4)a59 ± 5 (4)a82 ± 15 (4)a141 ± 15 (4)ab
GHNP 39.2 ± 1.3 (4)d,e13.8 ± 1.1 (4)a,b29.3 ± 1.3 (4)a4.3 ± 0.3 (4)c6.9 ± 0.3 (4)bc47 ± 6 (4)a79 ± 8 (4)a126 ± 13 (4)ab
ctl 38.9 ± 0.3 (4)a4.3 ± 0.3 (4)a24.6 ± 1.7 (4)a,b2.69 ± 0.06 (4)a9.2 ± 0.7 (4)a13 ± 3 (4)a60 ± 8 (4)a73 ± 6 (4)a
NP 35.8 ± 0.6 (4)b,c4.8 ± 0.4 (4)a,b29.5 ± 1.2 (3)a4.0 ± 0.3 (4)b7.7 ± 1.1 (3)a13.9 ± 1.9 (4)a56 ± 9 (4)a70 ± 8 (4)ab
SH 36.6 ± 0.6 (4)c,d4.6 ± 0.1 (4)b,c 2.6 ± 0.2 (4)a 10.7 ± 0.5 (4)a61 ± 7 (4)a72 ± 7 (4)b
SHNP  5.6 ± 0.3 (4)c29.0 ± 0.7 (4)a4.7 ± 0.4 (3)b6.4 ± 0.4 (3)a   
GH 36.6 ± 0.7 (4)c,a4.8 ± 0.2 (4)a,b22.8 ± 0.6 (4)b2.85 ± 0.1 (4)a8.00 ± 0.1 (4)a10 ± 1 (4)a55 ± 7 (4)a65 ± 7 (4)ab
GHNP 38.5 (1)4.4 ± 0.1 (3)a,b24.4 ± 1.6 (2)a,b4.5 ± 0.4 (2)b5.39 (1)9.7 (1)52 (1)61 (1)
ctl 30.1 ± 0.7 (4)a10.5 ± 1.6 (4)a29.7 ± 1.6 (4)a3.3 ± 0.3 (4)a9.3 ± 0.8 (4)a17.6 ± 1.9 (4)a95 ± 11 (4)a113 ± 9 (4)a
NP 27.4 ± 1.2 (4)b,c14.4 ± 0.7 (4)a,b29.0 ± 0.3 (4)a6.04 ± 0.05 (4)b4.80 ± 0.04 (4)b16.7 ± 1.7 (4)a85 ± 4 (4)a102 ± 3 (4)ab
SH 29.5 ± 0.5 (4)c,d18 ± 4 (4)b,c29 ± 2 (4)a4.67 (1)c7.21 (1)ab14.6 ± 1.1 (4)a70.3 ± 1.4 (4)a84.9 ± 1.8 (4)b
SHNP 26.7 ± 0.9 (4)b18 ± 1 (4)37.6 ± 0.7 (4)b6.74 (1)b5.38 (1)b15.6 ± 1.5 (4)a75.6 ± 1.3 (4)a91 ± 2 (4)b
GH 32.0 ± 1.4 (4)c,a15.8 ± 0.8 (4)a,b31.1 ± 1.5 (4)a3.1 ± 0.2 (4)a10.1 ± 0.4 (4)a17.2 ± 1.0 (4)a87 ± 5 (4)a104 ± 5 (4)ab
GHNP 31.6 ± 1.8 (4)d,e15.2 ± 1.2 (4)a,b34 ± 2 (3)a,b3.25 ± 0.2 (3)a10.0 ± 0.6 (2)a14 ± 3 (4)a80 ± 3 (4)a94 ± 3 (4)a,b

Treatment effects on foliar nitrogen were similar for all species except for Rubus growing in the NP plots (Tables 1, S2). Leaf N increased by 27 ± 5% in fertilized Betula and Eriophorum in the open and in shade but not in GH or GHNP plots (Table 1; > 0.2). Leaf phosphorus increased 43–133% in response to fertilization, except for Rubus GHNP, but with strong interaction by Species, Nutrient and Environment (Table S2). Glasshouse warming alone did not affect foliar P concentration in any species (> 0.5 in each species). Only one P measurement was available for SH Betula and SH Rubus leaves, each bulked from two experimental blocks. Nevertheless, both measurements were higher than P values for corresponding species in the control plots, implying that shade house conditions may have increased P-availability for these two species. Leaf N : P was < 10 in all treatments (Table 1), below the threshold of 12–14 associated with the transition from N– to P–limitation (Aerts & Chapin, 2000; Tessier & Raynal, 2003). Adding fertilizer reduced the N : P ratio, due to the larger increase in leaf P (Table 1).

Leaf carbohydrate response was similar for all species (Tables 1, S2). Foliar starch content in shaded plants was 11 ± 5% lower than in control plots (Table 1; = 0.04). Fertilization did not affect leaf starch contents (> 0.4 for the Nutrient term). Similarly, total sugar concentrations decreased 14 ± 6% in shaded vs open plots (= 0.03) whereas nutrient effects were not statistically significant (= 0.17). TNC followed the same trends as its component measurements but with increased statistical significance: TNC decreased 13 ± 4% in shade house leaves vs open plots (= 0.001; Table 1). A 6 ± 3% decrease in TNC in response to nutrient addition was marginally significant (= 0.06) and is weakly supported by the AIC analysis, with wi of 52% in support of the full additive model, S + N + E, making it twice as likely an explanation of the data as that of the next-best model, S + E (wi = 26%; Table S2).

Mitochondrial size and density, quantified and analysed for leaf sections sampled from the uncovered plots, differed significantly between species (P < 0.001, Table 2). Under fertilization, mitochondrial density increased in Betula and Eriophorum (P < 0.05 and 0.001, respectively). Concurrent decreases in mitochondrial size (P = 0.04 and 0.01 in Eriophorum and Rubus, respectively), however, resulted in comparable total average cross-sectional mitochondrial area in Eriophorum and Betula across treatments, and a decrease in this metric in Rubus under fertilization (Table 2).

Table 2. Mitochondrial size (= 30), density (= 45) and total average cross-sectional area of mitochondria per cell for the three species (Betula nana, Eriophorum vaginatum and Rubus chamaemorus) under control (ctl) and fertilization (NP) treatments
 TreatmentDensity (Number per 100 μm−2 cell)Size (μm2)Cross-sectional Areaa (μm2 mito μm−2 cell)
  1. a

    Cross-sectional area was calculated as the product of Density and Size. Uncertainty for this value was propagated according to Eqn 7. Values represented are mean ± SE.

Betula ctl12.1 ± 0.70.17 ± 0.012.03 ± 0.17
NP17.4 ± 1.10.12 ± 0.012.16 ± 0.22
Eriophorum ctl13.7 ± 1.30.19 ± 0.012.66 ± 0.29
NP20.5 ± 1.70.11 ± 0.012.28 ± 0.28
Rubus ctl7.5 ± 0.60.28 ± 0.032.10 ± 0.28
NP7.4 ± 0.60.19 ± 0.021.41 ± 0.19


Foliar R20 increased in response to fertilization for Betula and Rubus in open and SH plots but not in the glasshouses (Figs 1a,c, 3a). No clear treatment response is seen in Eriophorum (Fig. 1b), but interpretation is confounded by the missing SHNP data. As a result, AIC analysis found three models to be equally plausible, given the data (Table S2): S + N × E (wi = 0.31), S × N + E (wi = 0.26), and S + N + E (wi = 0.25). The first model discounts the apparent lack of increased R20 in Eriophorum NP, thus finding no interaction by species, but with an environmental interaction due to the reduced R20 in the glasshouses; the second model suggests that Eriophorum differed from the other two species in its reaction to nutrients but still finds R20 universally reduced in the glasshouses; the third model discounts all apparent interactions, implying instead that nutrient addition raised R20 in all environments, and that glasshouses lowered R20 equally in all species. In order to resolve the uncertainty regarding species interaction, we created a new variable, S2, that grouped the apparently similar Betula and Rubus responses, a method suggested by Crawley (2007). When compared with all previous models, the new model, S2 × N + E, was judged to be three-fold better than the other models (Table S2). ANOVA confirmed the significance of S2 × N (= 0.03) as well as the additive term for E (< 0.001). Based on this model, leaf R20 in Betula and Rubus plants grown at ambient temperatures increased 20–30% in response to fertilization. In all candidate models, R20 was reduced by 14 ± 4% across species in plants grown in the glasshouse, with no effect of fertilizer on any of the species sampled in the glasshouse (Fig. 1).

Figure 1.

Respiration at 20°C on a leaf-mass basis, R20 (a–c), and respiratory discrimination against 18O (d–f) in three species (mean ± SE): Betula nana, Eriophorum vaginatum and Rubus chamaemorus. Betula values are for whole-shoot tissue. Letters in the top margin of the respiration graph represent statistically distinct ‘Environment’ groups at the < 0.05 significance level; letters at the top of the discrimination graph represent statistically distinct treatment responses. An additive increase in R20 in response to fertilization was also statistically significant in the pooled data. The main text discusses possible interaction effects. Treatment codes are ctl (control), shade house (SH) and glasshouse (GH) with or without added fertilizer (NP). = 4 except for Eriophorum GHNP, which consists of a single measurement.

Respiratory electron partitioning

Initially, no treatment differences in respiratory discrimination were detected for Betula shoots or for Eriophorum or Rubus leaves (Fig. 1d–f; Table S2). Treatment effects for Betula, however, may have been masked by incubating stem and leaf tissue together: when leaves were incubated separately, discrimination of 23.9 ± 0.3‰ in fertilized plants was 1.6 ± 0.4‰ lower than in the control leaves (Fig. 2b; = 0.0006 in ANOVA with the corresponding Eriophorum and Rubus treatments). Discrimination in stems, 18.9 ± 0.8‰ and 19.8 ± 1.1‰ for NP and control, respectively, was lower and more variable than in leaves, resulting in nonsignificant differences (= 0.57). The reduced discrimination values and increased variability for stems is consistent with reports that measured discrimination is artificially decreased in dense tissue due to a pO2 gradient that depends on tissue thickness and R (Guy et al., 1989; Angert & Luz, 2001; Miller et al., 2011).

Figure 2.

Respiration (a) and respiratory discrimination against 18O (b) in Betula leaves and stems incubated separately (mean ± SE). Letters in the bars represent statistically distinct groups at the < 0.05 significance level. ctl, control plots; NP, fertilized, uncovered plots. = 4 for each bar.

In order to translate measured discrimination values into enzyme activity, Betula leaves were inhibited with selective respiratory inhibitors. Measured discrimination values for inhibited leaves were 26.4 ± 0.2‰, 21.0 ± 0.7‰ and 21.1 ± 1.3‰ for CP-inhibited, AP-inhibited and double-inhibited leaves, respectively (Table 3). Residual respiration in the double-inhibited leaves was 11% of the buffer-treated leaves, indicating effectiveness of the inhibitors. The measured difference of 5.3‰ between the endpoints, however, was considerably smaller than the mean reported difference of 10‰ across a variety of green plant tissues (Ribas-Carbo et al., 2005). COX and AOX activity was therefore estimated in two different ways. First τa, vcyt, and valt were computed using our measured endpoint values. Second, the three partitioning values were computed using the mean published endpoint values (Ribas-Carbo et al., 2005) to determine the extent of systematic error in case our anomalous endpoint values had been due to experimental circumstances rather than actual physiological differences (Table 3). Regardless of the assumed endpoint range, the increased R20 in response to fertilization can be attributed to a near-doubling of vcyt, from 9 to 17 nmol O2 g−1 DM s−1, whereas valt remained at c. 8 nmol O2 g−1 DM s−1 (Table 3).

Table 3. Two scenarios for calculating electron partitioning in Betula nana leaves
 D () τ a R v cyt v alt
  1. The first four rows are base on endpoints measured in this study; because of the uncertainty in these endpoints, a second set of rows uses the mean and SE of previously reported endpoint values across a variety of species (Ribas-Carbo et al., 2005). Quantities are: D, discrimination; τa, proportion of electrons partitioned through AOX; R, respiration (nmol O2 g−1 DM s−1); vcyt and valt represent the activity of the cytochrome and alternative pathways (nmol O2 g−1 DM s−1).

Measured Dc and Da
COX21 ± 0.70   
AOX26.4 ± 0.21   
ctl23.94 ± 0.160.54 ± 0.0717.50 ± 0.97.97 ± 1.279.53 ± 1.30
NP22.40 ± 0.30.26 ± 0.1123.90 ± 1.817.70 ± 2.986.20 ± 2.70
Published Dc and Da
COX19.9 ± 0.30   
AOX29.2 ± 0.51   
ctl23.94 ± 0.160.46 ± 0.0317.50 ± 0.99.4 ± 0.88.1 ± 0.7
NP22.40 ± 0.30.31 ± 0.0423.90 ± 1.816.6 ± 1.67.3 ± 1.1

Protein analysis

AOX/COX protein concentrations were measured in Betula and Eriophorum only, because the signal for Rubus was too faint for analysis (Fig. 3). No species or treatment interactions were supported (Table S2). Nutrient effects were only marginally supported, as indicated by order-dependent significance for N in the ANOVA tests: the Nutrient term is significant if N appears before E (= 0.008) but is not significant if E appears before N (= 0.14). The AIC analysis, which is order-independent, also finds that models excluding N (wi = 0.49) are only slightly more likely to explain the data the models that included N (wi = 0.39; Table S2. The more strongly supported Environment effect was due to a 59 ± 12% reduction in the signal ratio for the glasshouse treatments and a 40 ± 13% reduction in the shade houses relative to open-air plots (P ≤ 0.001 in both cases). Nutrient addition increased the AOX/COX signal by 15 ± 12%.

Figure 3.

Western blot analysis of (a, c) Betula leaves and (b, d) Eriophorum leaves (mean ± SE). Letters in the upper margin represent statistically distinct responses to Environment at the = 0.05 significance level. No statistically significant effects were found for nutrient addition, nor for interaction between any of the variables. See Fig. 2 for a key to the treatment codes. = 4 except for Eriophorum GHNP which consists of a single measurement.


The primary goal of this study was to determine whether plants growing in the field modulated respiratory efficiency in response to environmental change. The answer may depend on species. Only the most successful responder to fertilization, Betula, improved ATP synthesis efficiency by decreasing AP partitioning. Treatment effects on community composition and leaf traits provide evidence for changes in plant energy balance, which we hypothesized would drive AP/CP partitioning, and therefore provide a context for interpreting the respiratory observations.

Community composition was strongly affected by nutrient addition

Twenty years of nutrient, light and temperature manipulation altered production and species dominance in this Alaskan moist tussock tundra. As reported previously (Chapin et al., 1995; Bret-Harte et al., 2001), the control plots contained a diverse, low-statured assemblage of plants in which Eriophorum and Betula accounted for 35% of the relative cover (Fig. S2). As illustrated in Fig. S2 and in previous studies on similar plots, shade or warming alone had little observable effect on community composition (Chapin et al., 1995; Bret-Harte et al., 2001); fertilization with nitrogen and phosphorous fertilizer (NP), on the other hand, strongly influenced community composition and growth, depending on light and temperature. In open plots and in the glasshouses (Fig. S2d), existing Betula ramets increased their growth rate, branch length and branching until they overshadowed and crowded out the remaining vegetation (Bret-Harte et al., 2001) covering three quarters of those plots after 20 yr (Fig. S3). In the fertilized shade houses, however, Betula was not successful and, instead, Rubus formed dense mats that displaced much of the other vegetation (Fig. S2c). Eriophorum, which was the tallest plant in the control plots, grew so poorly in GHNP and SHNP, possibly due to increased herbivory (Gough et al., 2012), that we were not able to harvest sufficient plant tissue for the respiratory incubations, excepting one GHNP plot.

Leaf physical and chemical traits responded similarly in all species

Leaf DMC and SLA are simple proxies for leaf nutrient stress in many species (Hodgson et al., 2011). Light availability also affects these characteristics, as shaded leaves are often thinner and have less dry mass (Hodgson et al., 2011). Both shading and nutrient addition are therefore expected to reduce DMC and increase SLA, as was found in this study. Importantly, the experimental manipulations appear to have affected these measures equally in all three species (Table 1). Curiously, fertilization did not change DMC or SLA in the glasshouse-grown leaves, even in Eriophorum and Rubus, which would have experienced combined effects of nutrient addition and shading since because illumination in the understorey had been reduced by 90% (Bret-Harte et al., 2001). Nevertheless, a theme throughout this study is that the warming treatment dampened leaf-level effects of fertilization.

Leaf N and P status is a more direct indicator of cellular nutrient availability than is leaf structure but some caution is still warranted when interpreting these data, because leaf nutrient concentrations may reflect vacuolar rather than cytoplasmic concentrations (Lee & Ratcliffe, 1993). For example, the very high P uptake seen in this study (Table 1) has been described as luxury uptake (Chapin & Shaver, 1996) as it appears to be in excess of what the plant can utilize for growth. Additionally, increased growth may dilute foliar nutrient concentrations, as was seen in Betula grown in GHNP: no significant change in leaf N status was detected despite the large increase in biomass relative to controls (Fig. S2b,c). Nevertheless, adding soil nutrients generally improved leaf nutrient status, but as with the leaf structural measures, the improvements were similar across species. The one exception, a lack of increased leaf N in Rubus NP leaves, is counterbalanced by the fact that growth of these plants had clearly been enhanced (Fig. S2b). The much smaller N increase in the fertilized glasshouses may be due to increased dieback in the fertilized plots due to increased shading of the lower leaves by the dense canopy (Chapin & Shaver, 1996) and again illustrates the dampening effect of the glasshouse treatments on fertilization.

Nutrient-limited plants may accumulate carbohydrates that could have been allocated to growth had sufficient nutrients been available (Hermans et al., 2006). This relationship is not universal, however, and may depend on the extent of nutrient deficiency (Gonzalez-Meler et al., 2001; Noguchi & Terashima, 2006). Conversely, reduced light would be expected to lower carbohydrate status due to lower production. In addition, leaf carbohydrate status may regulate AP/CP activity directly (Azcón-Bieto et al., 1983; Florez-Sarasa et al., 2009). In this study, although carbohydrate concentrations decreased in response to shade and, somewhat more weakly, in response to nutrient addition, the responses did not differ by species (Table 1). In summary, whereas all leaf traits examined here were consistent with expected responses to changes in nutrient and light availability across all three species, no differences were found to explain their differential success rates at the community or whole-plant scale.

Mitochondrial organization in these species is dynamic and exhibits flexibility to changes in soil nutrient availability (Table 2). Previous studies have identified similar increases in leaf mitochondria number in plants grown under increased soil N and P and elevated CO2 treatments (Robertson et al., 1995; Griffin et al., 2001; Wang et al., 2004), suggesting a cross-taxa adaptive shift under increased energy demand that affects many regulatory pathways (Gonzalez-Meler et al., 2009; Leakey et al., 2009; Geigenberger et al., 2010). Nevertheless, changes in mitochondrial number were largely offset by changes in cross-sectional area, resulting in little net change in response to the treatments (Table 2). Although it is possible to correlate mitochondrial respiratory fluxes with functional mitochondria using confocal microscopy (Gomez-Casanovas et al., 2007), transmission electron microscopy used in this study may limit the potential for scaling. Despite this limitation, the data presented show a significant relationship in mitochondria organization and soil nutrient availability suggesting an adaptive organelle response that corresponds to increases in respiratory pathway regulation (Table 2).

Nutrient addition increased vcyt in Betula, but did not alter valt

Despite the lack of per-species differences in the leaf traits, we did find evidence for differences in their respiratory response: whereas the two ‘winners’ increased respiration rates in response to fertilization, leaf respiration in Eriophorum appears to have remained constant (Fig. 2a–c). Importantly, Betula, the most successful responder to fertilization, allocated the increase exclusively to the energy-efficient cytochrome pathway (Table 3), thus optimizing the benefit of increased nutrient intake. At the whole-plant level, Betula's success is attributed to a change in growth patterns, from producing mostly short shoots to producing many longer shoots with increased branching and increased leaf production (Bret-Harte et al., 2001) and by effective utilization of mycorrhiza for below-ground carbon transfer between plants (Deslippe & Simard, 2011). Our evidence suggests that as part of these changes, Betula also upregulated its metabolic machinery to support the increased growth both by increasing output and increasing efficiency. The respiratory response in Betula is similar to that found in a recent laboratory study of Arabidopsis thaliana leaves in which CP activity scaled with growth rate whereas AP activity remained nearly constant (Florez-Sarasa et al., 2007; Priault et al., 2007). In other studies, however, CP and AP activities scaled proportionally with increasing respiration (Florez-Sarasa et al., 2011) or varied by species (Gonzalez-Meler et al., 2001). Finally, we note that although we, of necessity, measured respiration in the dark, recent evidence indicates that plant respiration retains the metabolic influences of its light regime even after several hours in the dark (Florez-Sarasa et al., 2012).

Energy balance may explain the lack of partitioning response in Rubus

We found no evidence of a change in respiratory electron partitioning in Rubus despite its increased respiration and growth success in the NP and SHNP plots (Fig. S2). Rubus, however, is an understory species, which is why it outcompeted the more light-dependent Betula in the shade houses. Under the energy-balance hypothesis we have proposed, low-light conditions would lead to low AP activity, as energy availability is now limiting growth. In a comparison of shade and sun species, for example, Noguchi et al. (2001) found little to no AP activity in the shade species Alocasia odora (Lodd.). Thus, the lack of an electron-partitioning response to fertilization in Rubus may simply reflect an already low AP activity. In support of this hypothesis we note that SHAM inhibition did not lower the measured discrimination rate in Rubus leaves, even though residual respiration in doubly inhibited tissue was 11% of the controls (Table S1). Likewise, the AP capacity – determined as the difference in R between KCN- and doubly-inhibited tissue – was only 15% of total respiration in Rubus grown in the NP plots (Table S1).

Circumstantial evidence for respiratory acclimation

Warming resulted in a 15% reduction of R20 for all species grown in the glasshouse treatments (Fig. 2a–c), suggesting that respiration in these species had partially acclimated to the higher growth temperature. Had there been no acclimation, R20 should have been unaffected by growth temperature. On the other hand, had the leaves fully acclimated, respiration at growth temperature would have been the same in each treatment and R20 would have been reduced 25% in the GH treatments, based on the Q10 values and the 4°C difference in growth temperature. Therefore, acclimation appears to have been 60% efficient. This evidence of acclimation may be of relevance to ecosystem model developers, because changes in R based on instantaneous measurements, that is, Q10, may not accurately reflect long-term changes in respiration rates (Gifford, 2003). Another effect of increased temperature was a dampening of fertilization effects on R20 (Fig. 2a–c). A recent meta-analysis found that leaf R scaled strongly with N (Reich et al., 2008) and this trend is reflected in the current dataset (Table 1, Fig. S4). Thus, the lack of respiratory response to fertilization in GHNP can be explained by the trends in foliar N content, which may have remained constant due to dilution of the N through increased growth or due to increased dieback (Chapin & Shaver, 1996), as mentioned previously.

Plants grown in the light had higher AOX/COX concentrations

Changes in electron partitioning may be due to increased protein content as well as metabolic regulation (Atkin et al., 2005b). We had hypothesized, based on published laboratory studies, that nutrient stress would be associated with increased AOX abundance. In the two species we were able to assay, the Western blot analysis revealed an increase in AOX concentration at the highest light intensities, consistent with our energy-balance hypothesis as well as with published reports linking increased AP capacity with higher light conditions (Noguchi et al., 2001; Florez-Sarasa et al., 2011; Zhang et al., 2011). Thus, even though we did not detect extra AP activity during our respiratory incubations, which must be conducted in the dark, it does appear that Betula and Eriophorum had increased AP capacity when grown in high light. Zhang et al. (2011) found that the AP may interact directly with photosynthetic intermediates to prevent photoinhibition in the chloroplasts. It is possible, therefore, that the increased AOX in the plants grown in the uncovered plots reflects a need for increase AP activity in the light, when it is needed.

In conclusion, the evidence presented here partially supports the energy balance hypothesis for respiratory electron partitioning. Fertilization decreased τa in Betula nana as predicted, but the decrease was due to an increase in cytochrome pathway activity rather than a decrease in the alternative pathway. This ability to increase the more efficient respiratory pathway independently of the AP may have contributed to Betula's developmental success in response to fertilization. Rubus, which thrived in the shade, appears not to have decreased AP activity, perhaps because activity was already low. Eriophorum which was the least successful in the fertilized plots had not increased leaf respiration (nor electron partitioning) in response to fertilization. Thus, none of our field-collected data provide evidence for increased AP activity in response to nutrient stress. Nevertheless, in agreement with the energy balance hypothesis, species growing in higher-light environments contained higher relative concentrations of AOX protein, implying an increased need for AP activity when energy is more abundant.


This work was supported by a grant from the Marsden Fund of the Royal Society of New Zealand and the US National Science Foundation International Polar Year 2007 (IPY #07-32664), as well as scholarship grants for A. Kornfeld from Education New Zealand and the University of Canterbury. We are grateful to the Arctic LTER (in particular Gus Shaver) for allowing us to conduct research in these experimental plots. We would also like to thank Darren Smalley for help with the Western blots, Odhran O'Sullivan for the Q10 data, and Heather Greaves for the SLA data. Finally, we would like to thank three anonymous reviewers as well as the handling editor for advice that has improved the final presentation of this work.