Increasing atmospheric CO2 reduces metabolic and physiological differences between isoprene- and non-isoprene-emitting poplars


  • Danielle A. Way,

    Corresponding author
    1. Nicholas School of the Environment and Department of Biology, Duke University, Durham, NC, USA
    2. Department of Biology, Western University, London, ON, Canada
    • Author for correspondence:

      Danielle A. Way

      Tel: +1 519 661 2111 ext. 88734


    Search for more papers by this author
    • These authors contributed equally to this work.
  • Andrea Ghirardo,

    1. Research Unit Environmental Simulation, Institute of Biochemical Plant Pathology, Helmholtz Zentrum München, Neuherberg, Germany
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Basem Kanawati,

    1. Research Unit Biogeochemistry and Analytics, Helmholtz Zentrum München, Neuherberg, Germany
    Search for more papers by this author
  • Jürgen Esperschütz,

    1. Center of Life and Food Sciences Weihenstephan, Chair of Soil Ecology, Technische Universität München, Neuherberg, Germany
    2. Research Unit Environmental Genomics, Helmholtz Zentrum München, Neuherberg, Germany
    Search for more papers by this author
  • Russell K. Monson,

    1. School of Natural Resources and the Environment and Laboratory for Tree Ring Research, University of Arizona, Tucson, AZ, USA
    Search for more papers by this author
  • Robert B. Jackson,

    1. Nicholas School of the Environment and Department of Biology, Duke University, Durham, NC, USA
    Search for more papers by this author
  • Philippe Schmitt-Kopplin,

    1. Research Unit Biogeochemistry and Analytics, Helmholtz Zentrum München, Neuherberg, Germany
    Search for more papers by this author
  • Jörg-Peter Schnitzler

    1. Research Unit Environmental Simulation, Institute of Biochemical Plant Pathology, Helmholtz Zentrum München, Neuherberg, Germany
    Search for more papers by this author


  • Isoprene, a volatile organic compound produced by some plant species, enhances abiotic stress tolerance under current atmospheric CO2 concentrations, but its biosynthesis is negatively correlated with CO2 concentrations. We hypothesized that losing the capacity to produce isoprene would require stronger up-regulation of other stress tolerance mechanisms at low CO2 than at higher CO2 concentrations.
  • We compared metabolite profiles and physiological performance in poplars (Populus × canescens) with either wild-type or RNAi-suppressed isoprene emission capacity grown at pre-industrial low, current atmospheric, and future high CO2 concentrations (190, 390 and 590 ppm CO2, respectively).
  • Suppression of isoprene biosynthesis led to significant rearrangement of the leaf metabolome, increasing stress tolerance responses such as xanthophyll cycle pigment de-epoxidation and antioxidant levels, as well as altering lipid, carbon and nitrogen metabolism. Metabolic and physiological differences between isoprene-emitting and suppressed lines diminished as growth CO2 concentrations rose.
  • The CO2 dependence of our results indicates that the effects of isoprene biosynthesis are strongest at pre-industrial CO2 concentrations. Rising CO2 may reduce the beneficial effects of biogenic isoprene emission, with implications for species competition. This has potential consequences for future climate warming, as isoprene emitted from vegetation has strong effects on global atmospheric chemistry.


Atmospheric CO2 concentrations have varied considerably in geological time (Beerling & Royer, 2011). CO2 concentrations have increased from 280 ppm in the middle of the 19th Century to current values of c. 395 ppm (Leuenberger et al., 1992; Tans & Keeling, 2012). Anthropogenic fossil fuel use and land use change are expected to continue increasing atmospheric CO2 concentrations, with concentrations of up to 1020 ppm predicted for the year 2100 (Meehl et al., 2007). Low atmospheric CO2 concentrations during and succeeding the Last Glacial Maximum are thought to have imposed selective pressures on organisms, and acted as a key driver in the evolution of plant traits such as C4 photosynthesis (Ehleringer et al., 1991; Osborne & Sack, 2012; Sage et al., 2012). The higher atmospheric CO2 concentrations of the current time and those predicted for the future may negate the advantage of such traits, including that of C4 plants over their C3 competitors (Sage & Kubien, 2003).

While photosynthesis is directly affected by CO2 concentrations through substrate availability for the Calvin–Benson cycle, CO2 concentrations also impact other leaf physiological processes, such as isoprene emission rates (Wilkinson et al., 2009; Possell & Hewitt, 2011). Isoprene (2-methyl-1,3-butadiene) is a volatile organic chemical that is emitted by some, but not all, plant species and is synthesized in chloroplasts through the 2-C-methyl d-erythritol 4-phosphate (MEP) pathway (Harley et al., 1999; Monson et al., 2013). Isoprene biosynthesis is negatively correlated with atmospheric CO2 concentration (Rosenstiel et al., 2003; Wilkinson et al., 2009; Possell & Hewitt, 2011), and the multiple independent gains and losses of isoprene emission in the plant kingdom may be linked, in part, to swings in the geological history of atmospheric CO2 concentrations (Monson et al., 2013). Isoprene is one of the most abundant hydrocarbons emitted by plants, and these emissions play an important role in modifying atmospheric chemistry (Fuentes et al., 2000), particularly in extending the lifespan of methane (Poisson et al., 2000; Archibald et al., 2011) and in contributing to the formation of secondary organic aerosols (Kiendler-Scharr et al., 2012).

Although the exact biochemical or biophysical mechanism(s) is unknown, isoprene increases photosynthetic tolerance to the oxidative stresses most frequently produced during periods of high leaf heat loads, high photosynthetic photon flux densities, low soil water availability and low atmospheric CO2 concentrations (Sharkey & Singsaas, 1995; Behnke et al., 2007, 2010a; Vickers et al., 2009; Way et al., 2011). Isoprene increases the stability of photosynthetic processes associated with chloroplast thylakoids (Velikova et al., 2011) and reacts with reactive oxygen species (ROS; Jardine et al., 2012), with both mechanisms potentially providing increased tolerance of abiotic stresses. Photosynthesis is more susceptible to stress caused by high temperature and high photon flux densities at low atmospheric CO2 concentrations because photosynthetic carbon reduction rates are reduced: the reduced sink capacity of the Calvin–Benson cycle at low CO2 increases the need for alternative mechanisms, such as isoprene production, that can either dissipate excess excitation energy through nonphotochemical processes or remove oxidative radical species that result from excess energy in the photosynthetic apparatus (Behnke et al., 2010a). Because isoprene's effect is dose-dependent (Singsaas et al., 1997), isoprene-based photosynthetic stress tolerance is also greater at low CO2 than at high CO2 (Way et al., 2011). Any advantage in maintaining photosynthetic carbon gain at low CO2 may be offset by the cost of isoprene biosynthesis, which amounts to 10 moles of CO2, 24 moles of ATP and 14 moles of NADPH for each mole of isoprene produced (Niinemets et al., 1999). However, isoprene-emitting leaves grown at low CO2 had both higher net photosynthetic rates and lower dark respiration rates than non-emitting leaves, differences that more than compensated for the carbon cost of isoprene (Way et al., 2011).

If higher growth CO2 concentrations minimize phenotypic and metabolic differences between emitting and non-emitting plants, there could be reduced selection pressure on the trait of isoprene biosynthesis in a high-CO2 world. Because biogenic isoprene emissions impact climate (Poisson et al., 2000; Archibald et al., 2011; Kiendler-Scharr et al., 2012), shifts in the abundance of isoprene-emitting species or their emission rates across geologically relevant atmospheric CO2 concentrations could have significant effects on global atmospheric chemistry (Pacifico et al., 2012). With this context in mind, we therefore examined how low (190 ppmv), current ambient (390 ppmv), and elevated (590 ppmv) growth CO2 concentrations affected leaf metabolism and photosynthetic physiology in isoprene-emitting (wild-type (WT) and empty vector control (C)) poplar (Populus × canescens) lines and lines with suppressed isoprene biosynthesis (RA2 and RA22). We hypothesized that: (1) suppression of isoprene biosynthesis would cause the largest metabolic and physiological changes at pre-industrial CO2 concentrations, with smaller differences between emitting and non-emitting lines at current and future predicted CO2 concentrations; and (2) metabolic changes in suppressed lines compared with emitting lines would involve increases in abiotic stress tolerance mechanisms to compensate for the loss of isoprene-related stress tolerance.

Materials and Methods

Plant material

We used four lines of Populus × canescens (Aiton) Sm. (syn. Populus tremula × P. alba): two isoprene-emitting lines (WT and the control for the transgenic manipulation (C)) and two non-isoprene-emitting transgenic lines where isoprene synthase expression was silenced by RNA interference (RNAi); for more details of the plant lines, see Behnke et al. (2007). We chose two lines (RA2 and RA22), out of 10 transgenic lines, where isoprene emission rates were most suppressed (see Behnke et al., 2007). Additionally, three independent transgenic PcISPS:GFP lines of Populus × canescens (nine trees), in which the PcISPS (P. canescens isoprene synthase) promoter is fused to the enhanced green fluorescent protein (GFP) reporter gene (see details in Cinege et al., 2009), were used for imaging of ISPS promoter activity at different CO2 concentrations.

Detailed growth conditions of the experiment can be found in Way et al. (2011). Briefly, cuttings were established on misting benches in the Duke University Phytotron. Once roots had formed, they were planted in 1 : 1: 1 (v/v/v) sand : perlite : peat in 10 × 10 × 36 cm pots, and five or more plants per line (WT, C, RA2 and RA22) were moved into each of three growth chambers (Model M-13; Environmental Growth Chambers, Chagrin Falls, OH, USA). Chambers were set for low (190 ppm), ambient (390 ppm) or high (590 ppm) CO2 concentration, as measured with an infrared gas analyzer (LI-COR 6252; Li-Cor, Lincoln, NE, USA) every 2–5 min. Elevated CO2 was attained by injecting pure CO2 into the ambient airstream as needed, and low CO2 was achieved by scrubbing CO2 from the incoming air with soda lime. Treatments were rotated between chambers every 3 wk to minimize chamber effects. All chambers supplied 700 μmol photons m−2 s−1 at canopy height over a 16-h photoperiod, and 27 : 23°C day : night temperatures. Plants were fertilized weekly with half-strength Hoagland's solution. Measurements were made on plants that had been exposed to treatments for 10–13 wk; there were no obvious size differences between plants from the different lines.

Gas exchange

Gas exchange and isoprene emission rates were assessed at growth CO2 concentrations (190, 390 and 590 ppm) at 30°C leaf temperature, saturating light (1000 μmol photons m−2 s−1) and 50% relative humidity using a portable photosynthesis system (Li-6400; Li-Cor). A fraction of the outgoing cuvette air was diverted to a chemiluminescence-based fast isoprene sensor (Hills Scientific, Boulder, CO, USA; see Hills et al., 1991). Measurements were made on the ninth leaf from the top of five trees from each of the WT, C and suppressed lines.

Confocal scanning laser microscopy

Leaf cross-sections were sampled from the ninth or tenth leaf from the top of plants carrying GFP fused to the PcISPS promoter (Cinege et al., 2009). Three independent GFP lines from each growth CO2 concentration were analyzed, with three sections taken from each of the nine trees. Sections were examined using a confocal laser-scanning microscope (Zeiss LSM 510 upright confocal and LSM Image Browser software; Zeiss, Jena, Germany). Setting details for images are as described in Cinege et al. (2009). Images were divided into five cell layers (upper and lower palisade cell layers and upper, middle and lower spongy mesophyll layers) by manually tracing cell outlines on-screen (ImageJ; US National Institutes of Health, Bethesda, MD, USA). Chlorophyll autofluorescence and GFP fluorescence intensity were analyzed in each cell layer for each image (ImageJ).

Analysis of photosynthetic pigments

The ninth leaf from the tree apex was harvested from five plants per line per CO2 concentration between 11:00 and 12:20 h, immediately frozen in liquid N2 and kept at −80°C until analysis. Pigment extractions and analyses were performed as described in Behnke et al. (2007). Briefly, 50 mg of frozen leaf material was extracted for 10 min in darkness at room temperature with 1 ml of acetone, then centrifuged for 10 min at 15 000 g and 4°C. The pellet was re-extracted with 500 μl of acetone. Both supernatants were unified and pigments were measured with a HPLC system (Model 515 pump, 717 cooled autosampler; both from Waters, Milford, MA, USA) using a UV/visible diode-array detector at 440 nm wavelength (Model 2996 and 447; Waters). The same leaf samples were used in the nontargeted metabolite and fatty acid analyses.

Nontargeted metabolite analysis

Nontargeted metabolite analysis was performed using a Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR-MS; APEX Qe; Bruker, Bremen, Germany) equipped with a 12-Tesla superconducting magnet and an Apollo II electrospray (ESI) source (Bruker). Metabolite identification was achieved via the MassTRIX web site (; Suhre & Schmitt-Kopplin, 2008) by using KEGG/API ( Details can be found in Behnke et al. (2010b) with the following modification: mass spectra were acquired with longer time domain transients (two MWords instead of one in the previous study) to achieve higher resolution, obtaining a mass resolving power of 200 000 at m/z = 400, sufficient for biological samples. Ions were produced only in the negative ionization mode of electrospray and the masses were corrected against H+ loss. Peak mass intensities (threshold of 1e6) were exported to a peak list at a signal to noise (S/N) ratio of 2 and submitted for annotation to an automated database search using MassTRIX; the maximal error accepted was 3 ppm, the database chosen for annotation was KEGG with expanded lipids (HMDB, LipidMaps; v. 06-2009), and the organism selected was Populus trichocarpa.

Phospholipid fatty acid extraction

Phospholipid fatty acids (PLFAs) were extracted from frozen leaves, using a procedure modified after Cho et al. (1992). In brief, 50 mg of frozen leaf material was extracted with CHCl3 : CH3OH : 1 N HCl (40 : 80 : 1, v/v/v), after centrifugation at 5000 g, and subsequently washed with water and a solution of 0.5 M HCl in CH3OH : H2O (1 : 1, v/v). The lipid extract was further prepared for analysis after Zelles et al. (1995). After separating the phospholipid fraction from neutral lipids and glycolipids on a silica bonded phase column (SPE-SI 2 g 12 ml−1; Bond Elut; Agilent Technologies, Palo Alto, CA, USA), fatty acid methyl esters (FAMEs) were obtained after mild alkaline hydrolysis and prepared for GC separation using myristic FAME as an internal standard. FAMEs were measured using a 5973MSD GC-MS (Agilent Technologies) linked via a combustion unit to an isotope ratio mass spectrometer (DeltaPlus; Thermo Electron Cooperation, Bremen, Germany) and identified via established fatty acid libraries and characteristic retention times. Fatty acids are designated as the total number of C atoms followed by the number of double bonds and their location (ω) after the colon. Saturated straight-chain fatty acids are indicated by ‘n’.


Confocal microscopy data were analyzed with a two-way ANOVA using growth CO2 treatment and cell layer (JMP Pro 9; SAS Institute, Cary, NC, USA). There were no differences between the two emitting and two non-emitting lines in any gas exchange parameter across the CO2 treatments (two-way ANOVA and post hoc Tukey test (> 0.05)), so gas exchange data were analyzed with a two-way ANOVA based on isoprene emission capability and growth/measurement CO2 concentration (JMP Pro 9).

Metabolomic differences and similarities among samples were revealed using principal component analysis (PCA) and validated by ‘full cross-validation’ using the software package The Unscrambler (v. 8.0; CAMO A/S, Oslo, Norway). The peak list of FT-ICR-MS mass intensities was selected as the X variable (scaled with 1 SD−1 to have the same unit variance) and all samples (i.e. = 60) were used. Differences in metabolites among the four lines (WT, C, RA2 and RA22) were discovered by cluster analysis (Hierarchical Clustering Explorer (HCE) v3.0; Only masses showing opposite profiles between the two clustered WT/C and RA lines (resulting in 10 individual samples per CO2 concentration for emitters and for non-emitters; average linkage cluster method using Euclidean distance measurements to assess similarity/difference) with a 0.8 Pearson product-moment correlation coefficient threshold were taken into account.

To assess the significance of changes in metabolite profiles, a Student's t-test (< 0.05) was performed applying a false discovery rate (FDR) of 5% according to the Benjamini Hochberg modified correction (Benjamini & Hochberg, 1995; Benjamini et al., 2006; matlab R2011b; MathWorks, Natick, MA, USA). For each change in a metabolite concentration that was significant in at least one CO2 treatment, the log2 ratios of peak intensities between WT/C and RA2/RA22 were calculated (= 10), based on separation in the PCA analysis. The statistical significance of differences between WT/C and RA lines and CO2 concentrations in pigments and fatty acids was determined with two-way ANOVAs and post hoc Tukey tests (< 0.05; SigmaPlot 11.0; Systat Software Inc., San Jose, CA, USA).


Isoprene synthase promoter (PcISPS:GFP) activity, measured as GFP fluorescence intensity, was negatively correlated with growth CO2 concentration (Fig. 1a–f), and showed a cell-specific enhancement at low, but not higher, CO2 concentrations (Fig. 1d–f). PcISPS:GFP activity was significantly correlated with isoprene emission rates in WT/C plants grown in the three different growth CO2 concentrations (r2 = 0.994; = 0.050; Fig. 1g), with both the greatest promoter activity and the highest isoprene emission rates at the lowest CO2 concentration. In accordance with previous observations (Behnke et al., 2007; Way et al., 2011), isoprene emission rates were negligible and independent of CO2 concentration in RA lines, but negatively correlated with CO2 concentration in WT/C lines (Fig. 2d).

Figure 1.

(a–c) Poplar (Populus × canescens) representative leaf cross-sections from (a) low (190 ppm), (b) ambient (390 ppm), and (c) high (590 ppm) growth CO2 concentrations, showing red chlorophyll autofluorescence and green GFP fluorescence (indicating isoprene synthase (ISPS) promoter activity). (d–f) Chlorophyll autofluorescence intensity (closed symbols) and ISPS activity (measured by GFP intensity; open symbols) across leaf cell layers in isoprene-emitting lines grown at (d) low (190 ppm), (e) ambient (390 ppm), and (f) high (590 ppm) CO2 concentrations. Cell layers: 1, upper/adaxial palisade cells; 2, lower/abaxial palisade cells; 3, upper/adaxial spongy mesophyll cells; 4, middle spongy mesophyll cells; 5, lower/abaxial spongy mesophyll cells. Values are mean ± SE,= 3 trees per CO2 concentration, with three sections per tree. (g, h) Relationships between (g) mean leaf cross-sectional GFP fluorescence and leaf isoprene emission rate; (h) mean net CO2 assimilation rate and both mean leaf cross-sectional chlorophyll autofluorescence (circles) and leaf chlorophyll concentration (triangles). Values are mean ± SE; for chlorophyll and GFP fluorescence, = 3 trees, with three sections per tree; for isoprene emission and net CO2 assimilation rate, = 10 trees; for chlorophyll concentration, n = 5 leaves.

Figure 2.

Effect of growth CO2 concentration on: (a) net CO2 assimilation rate; (b) stomatal conductance; (c) the ratio of intercellular to ambient CO2 concentration (Ci/Ca); and (d) isoprene emission rate in isoprene-emitting (closed symbols) and isoprene emission suppressed (open symbols) poplar lines. Measurements were made at growth CO2 concentration, 30°C leaf temperature and 1000 μmol m−2 s−1 photosynthetic photon flux density (PPFD). Values are mean ± SE,= 10 trees. Asterisks indicate significant differences between lines (one-way ANOVA;< 0.05).

Chlorophyll autofluorescence was also greatest at low CO2 concentrations, peaking in the same cell layer as PcISPS:GFP activity (first palisade cells adjacent to the spongy parenchyma; Fig. 1). However, chlorophyll autofluorescence and net photosynthetic rates in the WT/C lines were negatively correlated (r2 = 0.686; = 0.035; Fig. 1h), with strong chlorophyll autofluorescence but low photosynthetic rates at the lowest CO2 concentration. WT/C poplars had higher net CO2 assimilation rates than RA lines at low (< 0.0001) and ambient CO2 (< 0.03), but these differences were not present at the highest CO2 concentration (Fig. 2a,b). Net photosynthetic rates increased and stomatal conductance decreased with measurement CO2 concentration, but there was no difference among lines in the ratio of intercellular to ambient CO2 concentrations (Ci/Ca) in plants grown in any of the three CO2 concentrations (> 0.05; Fig. 2a–c). Chlorophyll concentrations were positively correlated with chlorophyll autofluorescence across growth CO2 concentrations, and negatively correlated with net photosynthetic rates (r2 = 0.548; = 0.026; Fig. 1h), as net photosynthetic rates were constrained most at low atmospheric CO2 when CO2 substrate concentrations limit carbon uptake. PcISPS:GFP activity and chlorophyll autofluorescence were positively correlated across all CO2 concentrations and cell layers (r2 = 0.737; statistical difference: chlorophyll fluorescence: < 0.001; GFP fluorescence: = 0.067; Supporting Information Fig. S1).

Nontargeted metabolomic and PLFA analyses

For the nontargeted metabolite profile, leaves that developed at the lowest CO2 concentration formed a distinct group from leaves grown at ambient and elevated CO2 concentrations (Fig. 3). Results from a PCA showed that WT/C lines grouped separately from RA lines, with a stronger separation between lines in samples from the lowest CO2 concentration. Significant metabolic differences (P-values corrected to an FDR of 5%) between WT/C and RA lines were indicated for almost 1100 individual masses (Fig. S2), of which 60 (Table S1) could be identified through the KEGG database with the MassTRIX approach (Suhre & Schmitt-Kopplin, 2008). These identified metabolites were associated with numerous biochemical pathways, including carbohydrate, nitrogen, fatty acid, terpenoid, and ubiquinone metabolism (Fig. 4, Table S1). Metabolites associated with nucleotide and ubiquinone metabolism, and amino acids and other compounds related to nitrogen metabolism were enhanced in the leaves of RA plants. Sucrose and sorbitol, two key transport forms of carbohydrates, were also enhanced in RA leaves, but sugars related to the pentose phosphate cycle and glycolysis/gluconeogenesis (i.e. d-ribose, salicin 6-phosphate, phospho-d-glycerate and 6-phospho-d-gluconate) were enhanced in WT/C leaves compared with leaves from the RA lines. Calcitriol was reduced in the leaves of RA plants, especially those grown at the lowest CO2 concentration. Despite a stronger restructuring of the metabolome in non-isoprene emitting leaves at low CO2 than at higher CO2 concentrations, some metabolites (such as dodecanoid and dihydroxyhexadecanoic acids) were positively correlated with CO2 concentration (Fig. 4).

Figure 3.

Principal component analysis (PCA) of metabolomic data (x-axis = PC1 = 28%; y-axis = PC2 = 15%) from isoprene-emitting (wild-type (WT)/control (C)) and isoprene emission suppressed (RA2/RA22) poplar lines analyzed by nontarget Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR-MS). Colors represent different CO2 concentrations (red, 190 ppm; orange, 390 ppm; green, 590 ppm).

Figure 4.

Metabolomic differences between leaves from both isoprene-emitting lines (wild-type (WT)/control (C)) and both isoprene emission suppressed (RA) poplar lines grown at different CO2 concentrations (red, 190 ppm; orange, 390 ppm; green, 590 ppm), analyzed by nontarget Fourier transform ion cyclotron resonance mass spectrometer (FT-ICR-MS). Identifications were obtained by comparing the results with annotated masses of each compound in KEGG databases and relating them to the main metabolism (for multiple assignments, only one name is reported as ‘name sum’ + et al.; the complete list can be found in Supporting Information Table S1). An asterisk over a bar indicates a significant difference between WT/C and RA2/RA22 within that CO2 growth concentration (t-test; < 0.05; mean ± SE;= 10).

Across all pathways, concentrations of antioxidants (e.g. ascorbic acid and glutathione) were greater in the leaves of RA plants compared with WT/C plants, with larger differences between these lines at the lowest CO2 concentration (Fig. 4). Concentrations of thiamin monophosphate were also higher in RA leaves, compared with WT/C leaves, and greatly increased when plants were grown at the lowest CO2 concentration (Table S1). Growth at the lowest CO2 concentration stimulated photosynthetic pigment production in all lines, and de-epoxidized xanthophyll cycle pigment concentrations in leaves of RA lines, compared with higher CO2 concentrations (Table S2; Figs 1h, 5).

Figure 5.

(a) Photosynthetic pigment and α-tocopherol (α-toc) contents in isoprene-emitting (wild-type (WT)/control (C); solid bars) and isoprene emission suppressed plants (RA2/RA22; hatched bars) (red, 190 ppm; orange, 390 ppm; green, 590 ppm). Neo, neoxanthin; Vio, violaxanthin; Ant, antheraxanthin; Zea, zeaxanthin; Lut, lutein; β-Car, β-carotene; Chl a, chlorophyll a; Chl b, chlorophyll b; Sum Chl, chlorophyll + chlorophyll b; Sum Car, 1β-carotene + violaxanthin + antheraxanthin + zeaxanthin + neoxanthin; Sum Xant, antheraxanthin + violaxanthin + zeaxanthin. (b) Ratios of: xanthophyll/β-carotene (Xant/Car); β-carotene/chlorophyll (Car/Chl); chlorophyll a/chlorophyll b (Chl a/b); de-epoxidation (zeaxanthin + 0.5×antheraxanthin)/(zeaxanthin + antheraxanthin + violaxanthin). Asterisks indicate significant differences between WT/C and RA2/RA22 within a CO2 concentration (Tukey test; < 0.05; mean ± SE;= 10).

There were large differences in fatty acid metabolism between emitting and suppressed lines across the CO2 treatments (Fig. 4). Phosphatidic acid was up-regulated in leaves grown at the lowest CO2 concentration in RA lines (Fig. 4, Table S1). Because isoprene has been shown to reduce thylakoid trans-membrane leakage of protons (Velikova et al., 2011), we analyzed the PLFA content in more detail using GC-MS techniques. Leaves of RA lines had greater concentrations of unsaturated fatty acids (such as α-linoleic acid and α-linolenic acid) in their PLFAs than WT/C lines (Fig. 6). Suppression of isoprene emission significantly increased (< 0.05) the content of 16:1, 18:1 and 18:3 PLFA when plants were grown at low CO2 compared with higher CO2 concentrations (Table S2). Within individual fatty acid species, differences between lines were most pronounced when trees were grown at the lowest CO2 concentration (Fig. 6a). The percentage of unsaturated fatty acids and the double bond index in the PLFA were negatively correlated with growth CO2 concentration in RA plants (Fig. 6b), but there was no CO2-dependent trend in WT/C plants. There was also a CO2-dependent increase of oxylipin metabolites in RA leaves, including the lipid-based hormone signals jasmonic acid and methyl jasmonate (Fig. 4).

Figure 6.

Fatty acid content of the phospholipid fraction in C16–C22 saturated (:0) and unsaturated (:1,:2,:3) compounds in leaves from both poplar isoprene-emitting lines (wild-type (WT)/control (C); solid bars) and isoprene emission suppressed lines (RA2/RA22; hatched bars) (red, 190 ppm; orange, 390 ppm; green, 590 ppm; n16:0, palmitic acid; 16:1, palmitoleic acid; n17:0, heptadecanoic acid; n18:0, stearic acid; 18:1, oleic acid; 18:2, α-linoleic acid; 18:3, α-linolenic acid; n22:0, behenic acid). The double bond index (DBI) was calculated using the following equation (Liljenberg & Kates, 1985): inline image. Asterisks indicate significant differences between WT/C and RA2/RA22 within a CO2 concentration (Tukey test; < 0.05; mean ± SE;= 10).


Increasing CO2 concentrations diminished metabolic and physiological differences between isoprene- and non-isoprene-emitting poplars, in support of our first hypothesis. The effects of isoprene suppression on leaf metabolism were stronger at low CO2 concentration and gradually diminished as the atmospheric CO2 concentration was increased, in agreement with the known negative correlation between isoprene biosynthesis and CO2 concentration (Wilkinson et al., 2009; Possell & Hewitt, 2011; Fig. 7a). This divergence between the metabolic phenotypes of the lines highlights the extensive remodeling of cellular and physiological processes that occurs in the absence of isoprene biosynthesis, in particular at low CO2 concentration (Fig. 7b–d). In accordance with our second hypothesis, plants with suppressed isoprene biosynthesis grown at lower CO2 concentrations up-regulated compensatory stress tolerance mechanisms, as well as various compounds related to isoprene function or isoprenoid biosynthesis, but these effects were less apparent at high CO2 concentrations.

Figure 7.

(a) Metabolic differences between isoprene-emitting (wild-type (WT)/control (C)) and isoprene emission suppressed (RA2/RA22) poplar plants normalized to metabolic differences at 590 ppm. (b) Schema of metabolic shifts, and (c) their physiological and (d) functional impacts in isoprene emission suppressed lines compared with isoprene-emitting (WT/C) lines (*, see Way et al., 2011). ABA, abscisic acid; aa, amino acids.

Changes in leaf metabolism in response to suppression of isoprene biosynthesis are negatively correlated with CO2 concentration

The large metabolomic difference between RA lines and WT/C plants showed that suppression of isoprene biosynthesis has profound, CO2-dependent consequences for many more metabolites than may be expected on the basis of the loss of isoprene function. This is probably a consequence of a reduction in C fluxes through the MEP pathway and the negative correlation of isoprene biosynthesis with CO2 concentration. In poplar, isoprene biosynthesis is by far the main C sink of the plastidic MEP pathway (Sharkey & Yeh, 2001): C flux through the MEP pathway is c. 100-fold larger in isoprene-emitting than in non-emitting poplar and is negatively correlated to CO2 concentration (A. Ghirardo et al., unpublished data). The MEP pathway produces many key compounds involved in the synthesis and maintenance of the photosynthetic apparatus (e.g. carotenoids, plastoquinone and chlorophylls), antioxidant molecules (i.e. tochopherols; Munné-Bosch, 2005), phytohormones involved in plant development (giberellin) or stomatal closure (abscisic acid (ABA)), and the isoprenyl moiety for protein isoprenylation (Gerber et al., 2009). Thus, a perturbation of C fluxes when ISPS is suppressed explains the wide metabolic effects observed. Moreover, because there is cross-talk between the MEP pathway and the cytosolic isoprenoid mevalonic (MVA) pathway (Laule et al., 2003), perturbation of the MEP pathway could affect cytosolic metabolites such as phytosterols, dolichols, and farnesyl residues, with broad implications for cell membrane signal transduction, cellular sorting and cytoskeleton reorganization (Crowell, 2000).

When ISPS was suppressed, decreased demand for C entering the MEP pathway resulted in a strong metabolic shift across multiple pathways, with the strongest overall effects at low CO2, where isoprene synthesis is normally highest. A direct effect from isoprene suppression is the accumulation of the isoprene precursor dimethylallyl diphosphate (DMADP; Behnke et al., 2007) and possibly its isomer isopentenyl diphosphate (IDP), which might have affected the production of other MEP pathway products. In support of this, we found higher amounts of geranyl diphosphate (GDP) in suppressed lines, and up-regulation of another isoprenoid precursor, 2-C-methyl-d-erythritol 2,4-cyclo-diphosphate (MEcDP). However, xanthophyll and carotenoid concentrations were lower in suppressed lines relative to WT/C. The inverse CO2 dependences, whereby (1) concentrations of GDP and pigments are higher at low CO2 and (2) the concentration of MEcDP is lower at low CO2, indicate that different regulation mechanisms determine fluxes of metabolites within the MEP pathway and the subsequent pathway leading to higher terpenoids. Metabolic engineering of plastidic terpenoids impacts the formation of downstream products in transgenic Arabidopsis thaliana (Aharoni et al., 2003) and Mentha piperita L. (Mahmoud & Croteau, 2001). Our previous studies at ambient CO2 concentrations found that plastidic terpenoid concentrations were reduced (xanthophylls and carotenoids; Behnke et al., 2007) or slightly increased (carotenoids; Behnke et al., 2010b) in repressed lines compared with WT lines. The nontargeted metabolomic analysis showed that the impact of suppressing isoprene emission on terpenoid metabolism is generally smaller than expected, but is spread over different compounds, and has a strong CO2 dependence. Analysis of the expression of different plastidic terpenoid genes (1-deoxy-d-xylulose 5-phosphate synthase (DXS), 1-deoxy-d-xylulose 5-phosphate reductoisomerase (DXR) and phytoene synthase (PSY)) across emitting and non-emitting lines at ambient CO2 concentrations, however, showed no difference in transcript number throughout the growing season (Behnke et al., 2010b). Transgenic lines also showed lower stomatal conductance, possibly as a result of higher amounts of abscisic acid glucose ester compared with WT/C: foliar ABA, which promotes stomatal closure, is derived from the MEP pathway (Barta & Loreto, 2006).

It seems, therefore, that the regulation of C flux toward the synthesis of isoprene and related metabolites is controlled at different biochemical and post-transcriptional levels (Guevara-Garcia et al., 2005), all of which adjust terpenoid production. Earlier analysis of DXS activity, the putative controlling step of the MEP pathway (Lois et al., 1998; Estévez et al., 2001; Munoz-Bertomeu et al., 2006; Vallabhaneni & Wurtzel, 2009), showed lower enzyme activities in repressed poplar lines compared with WT lines (Ghirardo et al., 2010). Other regulatory controls might be under feedback mechanisms in the downstream part of the MEP pathway (Janowski et al., 1996), such as DMADP accumulation (Wolfertz et al., 2004). While the majority of our metabolic results show the negative CO2 dependence we predicted, unsurprisingly some metabolites show other correlations with CO2 that probably reflect alterations in metabolic fluxes or other processes not dealt with here. Consequently, future work should investigate the mechanisms controlling the regulation of the MEP pathway, as well as the effects of altering CO2 on the individually affected compounds.

Alternate stress tolerance mechanisms are up-regulated in suppressed lines in a CO2-dependent manner

The activity of PcISPS:GFP was regulated by growth CO2 treatment, indicating a role for atmospheric CO2 concentration as a signal for the transcriptional regulation of isoprene synthesis, along with other previously identified environmental factors such as light and temperature (Cinege et al., 2009). Isoprene synthase appears to be localized to the chloroplasts (Wildermuth & Fall, 1996; Schnitzler et al., 2005) and PcISPS:GFP activity was positively correlated with chlorophyll autofluorescence across all CO2 concentrations and cell layers. Thus, isoprene promoter activity (and isoprene synthesis) was greatest where chlorophyll content, and thus photosynthetic capacity, were also highest. ISPS promoter activity is highest in palisade mesophyll cells (Cinege et al., 2009), but the enhanced localization of ISPS promoter activity to this cell layer at our lowest CO2 concentration implies a greater reliance on the physiological advantages provided by isoprene biosynthesis in these cells. While low-CO2-grown plants also had the lowest net photosynthetic rates measured at their growth CO2 (because of limited CO2 substrate availability for photosynthesis), their high chlorophyll concentrations indicate that they would need to dissipate excitation energy toward alternative sinks, and thus support an enhanced need for stress tolerance mechanisms beyond that required for leaves that developed in higher CO2 concentrations.

While reduced stress tolerance has been noted before in studies of the repressed lines grown at ambient CO2 concentrations (Behnke et al., 2007, 2010a,b), our results highlight the role of changes in growth CO2 concentrations in exacerbating or diminishing stress-tolerance-related metabolic differences between isoprene-emitting and suppressed lines. Leaves from suppressed lines that developed at the lowest CO2 concentration had the highest concentrations of antioxidants, and of thiamin monophosphate, a precursor of thiamin biosynthesis that has been linked to increased oxidative stress tolerance (Tunc-Özdemir et al., 2009). Concentrations of de-epoxidized xanthophyll pigments were higher in suppressed lines grown at the lowest CO2 concentration, compared with isoprene-emitting poplars, but this difference disappeared when trees were grown at the highest CO2 concentration. A high de-epoxidation state is indicative of enhanced heat dissipation by nonphotochemical quenching in lines with suppressed isoprene emission (see Behnke et al., 2007; Way et al., 2011), and the de-epoxidation state increases in response to increases in the trans-thylakoid proton gradient that occur during insufficient dissipation of photosystem energy absorption (Holt et al., 2004; Demmig-Adams & Adams, 2006). Atmospheric CO2 concentrations also modulated the degree of unsaturation of fatty acids and the double bond index in the PLFAs, but only in non-isoprene-emitting lines. A shift toward unsaturation of the fatty acid component can provide stress tolerance to abiotic factors such as drought (e.g. Navari-Izzo et al., 2006) and high temperature (e.g. Gombos et al., 1994; Sato et al., 1996; Burgos et al., 2011), although it is often linked to low-temperature acclimation. These changes were accompanied by the occurrence of oxylipins when trees were grown at low CO2 concentrations. As oxylipins can be produced by the oxygenation of polyunsaturated fatty acids by free radicals (Méne-Saffrané et al. 2008; Durand et al., 2009), this may indicate increased oxidative damage in suppressed lines at the lowest CO2 concentrations. The pronounced rise in phosphatidic acid concentrations in leaves of the RA lines grown at the lowest CO2 concentration probably reflects a need to produce new membrane constituents under the higher stress conditions of a low-CO2 environment, as phosphatidic acid is a precursor for the biosynthesis of many lipids (in particular acylglycerol lipids) and acts as a signaling lipid to stimulate fatty acid biosynthesis (Eastmond et al., 2010; Hong et al., 2010).

Lastly, while we focus on abiotic stress, isoprene production can also deter herbivory (Laothawornkitkul et al., 2008). Jasmonic acid is important for up-regulating pathogen- and herbivore-defense pathways in leaves, and increased jasmonic acid concentrations in suppressed lines may reflect an increase in non-isoprene-mediated biotic defenses. Indeed, Behnke et al. (2012) observed a lower susceptibility of the same non-isoprene-emitting plants to the pathogenic fungus Pollaccia radiosa, compared with isoprene-emitting poplar lines.

Despite up-regulating multiple stress tolerance mechanisms, our earlier work showed that poplars with suppressed isoprene emissions still have a lower capacity for recovering from abiotic stress when grown at low CO2 concentrations (Way et al., 2011), demonstrating that this suite of changes does not fully compensate for the loss of isoprene biosynthesis. While the mechanism for isoprene-derived photosynthetic stress tolerance has been debated (Loreto & Schnitzler, 2010), in terms of whether it acts as an antioxidant (Vickers et al., 2009) or stabilizes thylakoid membranes (Velikova et al., 2011), our analysis shows that both antioxidant and membrane lipid metabolisms are significantly up-regulated when isoprene synthesis is suppressed, implying a role for both processes (see also Velikova et al., 2012). As other isoprene-emitting species show similar changes in emission rates with varying CO2 (Wilkinson et al., 2009), the conclusions drawn from our results may be broadly applicable. However, the extent to which any specific abiotic stress tolerance mechanisms are up-regulated by suppressing isoprene may depend on the extent to which a species relies on that mechanism.

We focus on the physiological and metabolic effects of suppressing isoprene biosynthesis under varying CO2 concentrations, but these results have implications for larger evolutionary and ecological questions. It has been proposed that isoprene biosynthesis evolved in geological periods of low CO2 (Way et al., 2011; Monson et al., 2013). Our findings are consistent with the potential for a strong selective pressure on isoprene biosynthesis at low CO2. On an ecological scale, differences in isoprene emission can alter competitive relationships between species (Lerdau, 2007), favoring isoprene-emitting phenotypes in environments with frequent abiotic stress. Thus, convergence of the metabolic phenotypes of isoprene-emitting and non-isoprene-emitting species may alter community dynamics as atmospheric CO2 concentrations rise, with significant implications for species composition in ecosystems such as tropical rain forests, which account for > 80% of global isoprene emissions (Müller et al., 2008). Reduced biogenic isoprene emissions in a future climate also have implications for changes in atmospheric chemistry that impact climate forcing, including secondary organic aerosol formation rates (Kiendler-Scharr et al., 2012) and atmospheric methane lifespans (Poisson et al., 2000; Archibald et al., 2011).


Differences in broad aspects of cellular metabolism between isoprene-emitting and non-isoprene-emitting plants diminished as the CO2 concentration was increased. The compensatory increases in multiple stress tolerance pathways in non-emitting leaves emphasize the critical role of isoprene in stress tolerance when plants are grown at low CO2 concentrations. As CO2 concentrations continue to increase, the diminishing metabolic and physiological differences between leaves that do and do not emit isoprene imply that the benefit of producing isoprene for abiotic stress tolerance also diminishes, which could lead to reduced selection for the maintenance of isoprene synthase in a future high-CO2 atmosphere.


The authors thank Will Cook and the Duke Phytotron staff for their assistance in growing the plants and the biostatistician Hagen Scherb for useful discussions on statistical approaches. D.A.W. was supported by the Natural Sciences and Engineering Research Council of Canada, and funding from the US Department of Agriculture (#2011-67003-30222), the US Department of Energy (#DE-SC0006967) and the US-Israeli Binational Science Foundation (#2010320). A.G., R.K.M. and J.P.S. were supported by a grant from the Human Frontier Science Programme (HFSP); R.B.J. acknowledges support from the US Department of Energy (#DE-FG02-95ER62083).