Gene expression patterns of trembling aspen trees following long-term exposure to interacting elevated CO2 and tropospheric O3


Author for correspondence: Gopi K. Podila Tel: +1 256 824 6263 Fax: +1 256 824 6305 Email:


  • • Expression of 4600 poplar expressed sequence tags (ESTs) was studied over the 2001–2002 growing seasons using trees of the moderately ozone (O3)-tolerant trembling aspen (Populus tremuloides) clone 216 exposed to elevated CO2 and/or O3 for their entire 5-yr life history.
  • • Based on replication of the experiment in years 2001 and 2002, 238 genes showed qualitatively similar expression in at least one treatment and were retained for analysis. Of these 238 genes, 185 were significantly regulated (1.5-fold) from one year to the other in at least one treatment studied. Less than 1% of the genes were regulated 2-fold or more.
  • • In the elevated CO2 treatment, relatively small numbers of genes were up-regulated, whereas in the O3 treatment, higher expression of many signaling and defense-related genes and lower expression of several photosynthesis and energy-related genes were observed. Senescence-associated genes (SAGs) and genes involved in the flavanoid pathway were also up-regulated under O3, with or without CO2 treatment. Interestingly, the combined treatment of CO2 plus O3 resulted in the differential expression of genes that were not up-regulated with individual gas treatments.
  • • This study represents the first investigation into gene expression following long-term exposure of trees to the interacting effects of elevated CO2 and O3 under field conditions. Patterns of gene-specific regulation described in this study correlated with previously published physiological responses of aspen clone 216.


Global increases in air pollutants are major threats to the functioning, structure and diversity of natural and seminatural ecosystems (Bobbink, 1998). Concentrations of the greenhouse gases carbon dioxide (CO2) (Keeling et al., 1995) and ozone (O3) (Fowler et al., 1999; Ryerson et al., 2001) are increasing rapidly. Elevated atmospheric CO2 in short-term exposures has been shown to increase photosynthesis (Drake et al., 1997; Will & Ceulemans, 1997; Tjoelker et al., 1998a; Noormets et al., 2001a,b), decrease respiration (Volin & Reich, 1996), and stimulate above-ground (Norby et al., 1999) and below-ground (King et al., 2001) growth. Trees grown under elevated CO2 generally have lower nitrogen concentrations in their foliage, lower rubisco concentrations (Moore et al., 1999), and altered defense compounds (Lindroth et al., 1993, 1997; Wustman et al., 2001). In long-term exposures, the effects of CO2 are often decreased (Rey & Jarvis, 1998; Tissue et al., 1999) as a result of reduced sink strength (Gesch et al., 1998) or nutrient-limited habitats (Bryant et al., 1998; Oren et al., 2001). Elevated CO2 has been reported to decrease the concentration of antioxidants such as glutathione and ascorbate (Schwanz et al., 1996; Polle & Pell, 1999; Wustman et al., 2001), but no effect of elevated CO2 on antioxidant transcripts has also been reported (Schwanz & Polle, 1998; Sehmer et al., 1998).

Tropospheric O3 is one of the most ubiquitous and damaging phytotoxins known (Broadmeadow, 1998). Ozone affects plant growth by reducing stomatal conductance, degrading rubisco and chlorophyll, decreasing photosynthesis and increasing photorespiration, inducing premature senescence and delaying bud break (Landry & Pell, 1993; Volin & Reich, 1996; Nali et al., 1998). Ozone enters mesophyll tissue through stomata and produces free hydroxyl radicals (OH), superoxide anions (O2 · –) and hydrogen peroxide (H2O2). This results in peroxidation and denaturation of cellular membranes (Wellburn & Wellburn, 1996; Ederli et al., 1997; Pell et al., 1997; Loreto & Velikova, 2001), leading to an accelerated foliar senescence through induction of ethylene, and changes in specific mRNAs (Glick et al., 1995) as reported in many species (Nie et al., 1993; Pell et al., 1997; Miller et al., 1999). Ozone induces hypersensitive stress responses (HRs) leading to apoptosis, which is coordinated by the signaling pathways of ethylene and jasmonic acid (Rao & Davis, 1999; Rao et al., 2000). Reactive oxygen species (ROSs) induced by O3 are scavenged in different subcellular compartments by antioxidant enzymes. Therefore, plants undergo numerous adjustments to counteract the effects of stress by regulating a number of genes involved in stress responses (Reymond et al., 2000; Cheong et al., 2002). Several studies have shown that O3 tends to increase expression of defense response genes such as glutathione S-transferase (GST), phenylalanine ammonia-lyase (PAL), and a cytosolic copper/zinc (Cu/Zn) superoxide dismutase (SOD) (Akkapeddi et al., 1999; Wustman et al., 2001). Photosynthetic protein mRNAs (Conklin & Last, 1995) and polyamines (Scalet et al., 1994) are also commonly up-regulated by O3. Ozone alters the thylakoid and chloroplast membranes in turn, decreasing photosynthetic capacity (Tognini et al., 1997) and weakening plasma membranes (Tokarska-Schlattner et al., 1997).

In the future, elevated tropospheric O3 will co-occur in the atmosphere with elevated atmospheric CO2 over large areas of the world's forests. Many studies have shown that elevated CO2 reduces the adverse effects of elevated O3 (Volin & Reich, 1996; Booker et al., 1997; Grams et al., 1999), while other studies have shown that elevated CO2 exacerbates the negative effects of elevated O3 (Kull et al., 1996; Lutz et al., 2000; McDonald et al., 2002; Percy et al., 2002). Very little is known about the impacts of these two greenhouse gases in combination on gene expression. However, Wustman et al. (2001) reported that CO2 did not ameliorate the harmful effects of O3. In aspen clones exposed to the co-occurring gases, increased damage to chloroplasts and suppression of the ascorbate-glutathione pathway were found (Wustman et al., 2001).

The impacts of elevated tropospheric O3 and elevated atmospheric CO2, alone or in combination, are being studied at the Aspen free-air carbon dioxide enrichment (FACE) project near Rhinelander, Wisconsin, USA. At this facility (Dickson et al., 2000), we have been exposing an aggrading northern forest ecosystem consisting of trembling aspen (Populus tremuloides Michx.), paper birch (Betula papyrifera Marsh.) and sugar maple (Acer saccharum Marsh.) to elevated CO2 and O3, alone and in combination, for the past 5 yr (Karnosky et al., 2003). For this study, we examined aspen clone 216, which is moderately O3-tolerant (Isebrands et al., 2001). Most previous studies of gene expression have been carried out on samples collected from the glasshouse or chamber with short-term exposures to either elevated CO2 or elevated O3 alone. In our study, we collected samples from trees that had grown under elevated CO2 and O3 over their entire 5-yr life histories.

The objective of this study was to examine the effects of elevated CO2 and/or O3 on global gene expression in aspen. Furthermore, we sought to detect linkages between gene expression and documented physiological responses following long-term exposure to elevated CO2 and O3, alone or in combination. We have shown that global gene expression analyses can be performed on field-grown trees and that these data can be linked to whole-tree responses.

Materials and Methods

The FACE experiment and biological materials

The Aspen FACE facility is located at the United States Department of Agriculture (USDA) Forest Service, Harshaw experimental farm near Rhinelander, WI, USA (Karnosky et al., 1999, 2003; Dickson et al., 2000). The experimental site consists of four rings each of control (ambient air; CO2 concentration 360 ppm), elevated CO2 (560 ppm), elevated O3 (1.5 times ambient) and elevated CO2 plus elevated O3 conditions in triplicate rings of 30-m diameter each. The eastern half of each ring consists of a randomized mixture of two-tree plots of five aspen clones including the relatively O3-tolerant clone 216 (Isebrands et al., 2001) used in this study. The remaining half of each ring is divided into two sections with either alternating paper birch and aspen or alternating sugar maple and aspen. Detailed information about the planting design, location and gas exposure in the Aspen FACE study is given in Karnosky et al. (1998, 1999, 2002) and Dickson et al. (2000).

Young, expanding aspen leaves of leaf plastochron index (LPI) 1–5 (Larson & Isebrands, 1971) were collected in August of 2001 and 2002 from the O3-tolerant clone 216 from the Aspen FACE facility. Leaves were harvested from 10:00 to 12:00 hours and samples were pooled from six trees from each of the three replicate rings from the FACE project. Leaves were wrapped in aluminum foil and immediately placed in liquid nitrogen before being taken to the laboratory, where they were transferred to −80°C storage until further use.

Generation of cDNA array

A cDNA array was designed from selected expressed sequence tags (ESTs) from adventitious roots (library R, 3984 ESTs) and leaves (library F, 486 ESTs) described in Kohler et al. (2003). The selected ESTs corresponded to single sequences or to ESTs representative of contigs of at least four ESTs (Kohler et al., 2003). Assembly of the individual ESTs into groups of tentative consensus sequences representing unique transcripts was performed using the contig routine (80% identity over a 40-nt length) of sequencher (version 3.1.1 for Macintosh; Gene Codes Corporation, Ann Arbor, MI, USA). A set of negative control spots corresponding to the cloning vector (four replicates of a pTriplex plasmid sequence and 12 replicates of a pTriplex polylinker sequence; BD Biosciences, Palo Alto, CA, USA), 17 replicates of a human desmine cDNA (kindly provided by T. Desprez, INRA Versailles, France), and 25 and 29 cDNAs encoding different genes from, respectively, fungal cDNA libraries of Pisolithus tinctorius and Laccaria bicolor (Peter et al., 2004) were also placed on the array along with 35 spot locations without nucleic material (water). All cDNAs from the root and leaf libraries were amplified from overnight culture of bacteria by polymerase chain reaction (PCR) and purified on the Multiscreen PCR system (Millipore, Molsheim, France), and the purity and length of all PCR products were checked by agarose gel electrophoresis. cDNA inserts (30–50 ng µl−1) were single-spotted onto positively charged nylon membranes by using the BioGrid arrayer (BioRobotics, Cambridge, UK) according to the array manufacturer's instructions (Eurogentec, Saraing, Belgium). The 0.4-µm pins of the 384-pin gadget deposited 100 nl of each PCR product and the final density of spots on the array was about 66 cDNA cm−2. Hybridization with radiolabeled probes corresponding to the plasmid polylinker of the plasmid vector used for construction of R and F libraries was routinely performed to ensure quality of cDNA arraying on filters.

RNA isolation, probe synthesis and hybridization

Total RNA was extracted from pooled leaf samples collected from six trees in each treatment, using a modified Chang et al. (1993) method. Polyvinylpolypyrrolidone (PVPP insoluble) was added to increase the purity of the RNA yield. Complex cDNA probes were then prepared by reverse transcription using Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA, USA) and the SMART-PCR cDNA synthesis kit (BD Biosciences) and hybridized to cDNA arrays as described in Duplessis et al. (2005). Different sets of arrays were used for hybridization of year-replicate samples and of samples from different conditions. Pooling of leaf samples from different trees from each FACE ring helped to minimize the variation among individual samples, arrays and probes.

Data analysis

The 16-bit TIFF images obtained with the phosphorimager imaging system were imported into the x-dot reader program (version 2.0; COSE, Paris, France) and quantified. Detection and quantification of the 4600 signals representing hybridized DNA (each cDNA plus appropriate controls duplicated) were performed using the ‘volume quantification’ method. Each spot was defined by automatic grid positioning over the array image and the average pixel intensity of each spot was determined. Net signal was determined by subtraction of global background measured all over the array from the intensity for each spot. Spots deemed not suitable for accurate quantification because of array artifacts were flagged and excluded from further analysis. The data table generated by x-dot reader, containing the intensity of each spot, was then exported to the Excel X:mac worksheet program (Microsoft Corporation, Redmond, WA, USA) for further manipulation. Spots that had an intensity of less than twice that of the background were flagged as undetectable and had their intensity raised to a minimum threshold value of 0.1 to avoid spurious expression level ratios at the bottom of the spot intensity range.

To take account of experimental variations in the specific activity of the cDNA probe preparations or in exposure time that might alter the signal intensity, the raw data obtained from different hybridizations were normalized (scaled) by dividing the intensity of each spot by the average of the intensities of all the spots present on the filter, to obtain a centered, normalized value (Eisen et al., 1998). Then we calculated a ‘flag limit’ corresponding to the remaining intensity measured at the locations of empty spots (water) on nylon arrays. This limit corresponded to the sum of the mean and the standard error calculated from signals measured at the 35 empty spot locations. We flagged and then separated from the data set all the spots that showed intensity lower than 2-fold the ‘flag limit’ in all conditions (both years 2001 and 2002 in control, CO2, O3 and combined CO2/O3 treatments). This filtering process allowed separation of clones close to the background noise level on the array from the analysis. The Microsoft Excel software was finally used to calculate the expression level ratios for remaining genes on arrays for all treatments (O3/control; CO2/control; combined CO2 + O3/control) in each year-replicate (2001 and 2002).

Data quality assessment was performed using analysis of variance (t-test) and a Bayesian statistical framework implemented in the Cyber-T web interface ( (Long et al., 2001; Baldi & Hatfield, 2002). Based on the statistical analysis, a gene was considered significantly up- or down-regulated if the t-test P-value was lower than 0.05 in at least one treatment (CO2 or O3 or combined CO2/O3) and the mean ratio associated with this is above a 1.5-fold limit. For the final analysis, the fold changes of gene expression in year-replicates were averaged and genes with expression falling between 0 and 1 were multiplied by −1 and inverted to facilitate their interpretation. Microsoft Excel spreadsheets with the remaining genes are available in Table S1 (available online as supplementary material) and at the website Excel spreadsheets with raw data and associated statistical results are also available at the same locations (CO2 vs control, Table S2; O3 vs control, Table S3; CO2 + O3 vs control, Table S4). Cyber-T includes a computational method (Posterior Probability of Differential Expression (PPDE)) for estimating experiment-wide false-positive and false-negative results based on the modeling of P-value distributions (Baldi & Hatfield, 2002), and results for these estimations are given in the two last columns of Tables S3 and S4. These columns are missing in Table S2 (CO2 conditions), where the PPDE calculation was impossible because of the data distribution in CO2 treatment replicate data sets.

Fold changes were supplied to the genecluster 2 program ( to identify groups of genes having similar patterns of regulation under the different treatments (Tamayo et al., 1999). In genecluster 2, the numbers of self-organizing map (SOM) rows and SOM columns were set to four and four, respectively, to explicitly build a SOM with 16 clusters (see Fig. 2 below). The number of iterations was set to 500 000 and the default parameters were used for the other settings. We only considered clusters with genes and removed those that remained empty (see Fig. 2 below). The calculation allowed us to derive genes with similar expression patterns without regard to the magnitude of the ratios. All descriptions of biological materials and procedures and raw data will be posted on the mycor website ( through public access to a BASE DB in construction which will comply with all MIAME (Minimum Information About a Microarray Experiment) standards set for microarray data (Brazma et al., 2001).

Figure 2.

Self-organizing map (SOM) clusters of expression profiles measured in leaves of aspen (Populus tremuloides) grown under elevated CO2, elevated O3 and combined elevated CO2 and O3 concentrations for genes that showed significant regulation of gene expression in at least one treatment when compared to the control conditions. Each graph displays the mean pattern of expression of the transcripts in that cluster (lines with solid circles) and the standard deviation of average expression (lines with no solid circles). The number of transcripts [expressed sequence tags (ESTs)] in each cluster is given in each SOM. The x-axis represents normalized gene expression and the y-axis the ratios measured in the three treatments compared with control conditions. The ESTs in each cluster are identified in Table S1 and mentioned in Table 2. Cluster number and number of genes in that cluster are given in each panel.

RNA blot analyses

For RNA blot analyses, electrophoresis under denaturing conditions was performed with 1.2% agarose containing 0.7 m formaldehyde (Lehrach et al., 1977) with RNA sampled in the year 2002. Gels were stained with ethidium bromide and blotted on nylon membranes (Hybond-N+; Amersham Pharmacia Biotech, Piscataway, NJ, USA) as described by the manufacturer. Hybridization was carried out as recommended by Amersham Pharmacia Biotech and signals were revealed by autoradiography using the Bio-Rad FX phosphorimager (Bio-Rad, Hercules, CA, USA).


FACE project and experimental design

In order to investigate the effects of elevated tropospheric O3 and elevated atmospheric CO2, alone or in combination, at the molecular level, we have studied gene expression in trembling aspen exposed to these treatments in the FACE project near Rhinelander, Wisconsin, USA. In this facility, among other northern forest trees, trembling aspens (P. tremuloides) were exposed to elevated CO2 and O3, alone and in combination, for the 5 yr before this study (Karnosky et al., 2003). In this study, we particularly examined the aspen clone 216, which is moderately O3-tolerant (Isebrands et al., 2001). Samples were taken in the years 2001 and 2002 from LPI 1–5 from six trees in each of the three replicate rings of the FACE project and pooled by treatments in each year. The experimental design for this gene expression study consisted of two ‘year-replicates’ of control conditions and the three considered treatments (elevated CO2, elevated O3 and combined elevated CO2 and O3).

Gene expression profiles in aspen leaves

To examine the gene activity changes associated with the long-term effects of exposure of aspen leaves to elevated CO2 and O3, alone or in combination, we performed a large-scale expression analysis using a nylon cDNA array. Radioactive [33P] cDNA probes were prepared from replicate sets (years 2001 and 2002) of RNA extracted from aspen leaves exposed to elevated concentrations of CO2 and O3, as shown in Table 1. Radiolabeled probes were incubated with 4608-element nylon cDNA arrays. These cDNAs correspond to Populus trichocarpa Torr. and Gray ×P. deltoides Bartr. ESTs from root (3984 ESTs) and leaf (486 ESTs) libraries previously described (Kohler et al., 2003). These clones encompass a large array of metabolic processes and response elements to biotic and/or abiotic stresses. In order to avoid as much as possible nonspecific hybridization signals between labeled aspen cDNA probes and poplar hybrid cDNA targets on arrays, our procedure included stringent washing conditions after hybridizations (Duplessis et al., 2005).

Table 1.  Summary of treatment exposures for the five years in which aspen (Populus tremuloides) trees were exposed to elevated CO2 and O3, alone and in combination, at the Aspen FACE facility
  • 1

    Daytime average (ppm).

  • 2

    Daytime mean concentration (ppb).

  • 3

    Sum 0 (ppmh).

CO21 (control and + O3)360360350365370
CO21 (+ CO2 and + CO2 + O3)530548545505504
O32 (control and + CO2) 34.6 36.9 34.5 36.6 33.1
O32 (+ O3 and + CO2 + O3) 54.5 51.7 48.1 52.8 49.0
O33 seasonal dose (control and + CO2) 59.1 62.8 55.8 61.6 50.6
O33 seasonal dose (+ O3 and + CO2 + O3) 97.8 89 79.9 90.7 80.8
Seasonal exposure duration (d)166143139143138

To test the quality of reproducibility between replicates sampled in the years 2001 and 2002, we hybridized sets of cDNA arrays with radioactive cDNAs prepared from RNA extracted during the years 2001 and 2002 and compared expression levels (normalized) for each treatment (Fig. 1). The Pearson correlation coefficient of 0.92 calculated for the control conditions indicated reasonably good reproducibility between the years 2001 and 2002. However, Pearson correlation coefficients measured for the treatments with elevated CO2 and O3, alone or in combination, were lower, from 0.67 (elevated O3) to 0.88 (combined elevated CO2 and O3), indicating less reproducible levels of expression for a large proportion of the genes plotted. This low correlation was expected for samples from trees grown in field conditions.

Figure 1.

Scatter plots comparing the signal intensities of pairs of arrays hybridized with two sets of cDNA probes prepared from RNA extracts from two year-replicates (2001 on the x-axis and 2002 on the y-axis). (a) Normalized mRNA abundance of 801 expressed sequence tags (ESTs) measured in leaves of aspen (Populus tremuloides) grown in control conditions. (b) Normalized mRNA abundance of 801 ESTs measured in leaves of aspen grown under elevated CO2 concentration. (c) Normalized mRNA abundance of 801 ESTs measured in leaves of aspen grown under elevated O3 concentration. (d) Normalized mRNA abundance of 801 ESTs measured in leaves of aspen grown under combined elevated CO2 and O3 concentrations. All axes show a log10 scale and arbitrary units. On each scatter plot, dotted lines indicate a 2-fold difference between year-replicates and the 1 : 1 ratio line. A linear regression equation and correlation factor r2 are given for each year-duplicate comparison.

We applied a filter to our data set based on the intensity of the background signal measured at empty spot locations on arrays (nonspecific hybridization). We flagged clones that were not above this limit in all treatments and in control conditions and separated them from the analysis. This flag limit was stringent enough to allow us to eliminate ESTs from the data set associated with low signal values for various reasons (e.g. sequence divergence between poplar species or tissue specificity of cDNA targets). This resulted in a final data set of 801 poplar ESTs after filtering. This relatively small number of ESTs, when compared with the 4608 initial ESTs spotted on the cDNA array, might be an expression of the sequence divergence between the two poplar species used in this study. However, a recent analysis of ESTs from different species showed little variation at sequence level between cDNAs from different poplar species (Sterky et al., 2004). Similar findings were obtained during heterologous array analyses between spruce and pine (van Zyl et al., 2002; Stasolla et al., 2004). The subset number of ESTs eliminated from the present data set may reflect a combined effect of the specificity of an important fraction of root cDNA targets on the array during hybridization with leaf cDNA probes and wash stringency after hybridization. Several experiments performed with the same poplar cDNA array hybridized with probes extracted from both leaf and root tissues have shown heterogeneity in signal intensities based on tissue specificity (A. Kohler, INRA, Nancy, France, pers. comm.). Moreover, the stringency during washes following hybridization could also have an important impact on the mean signal measured on a cDNA array (van Zyl et al., 2002). In the present study, this point is underlined by the percentages of root and leaf ESTs that survived the filtering step (12% and 66%, respectively) in the final data set. From among the 801 ESTs that were above the flag limit, we distinguished genes that were significantly regulated in at least one treatment between replicates sampled in the years 2001 and 2002 using analysis of variance (t-test) and the Bayesian statistical framework implemented in the Cyber-T web interface (ln P-value = 0.05; Tables S2, S3 and S4) (Long et al., 2001; Baldi & Hatfield, 2002). We then determined the level of regulation of gene expression by dividing the normalized mRNA concentration measured in the three treatments by the normalized mRNA concentration measured in control conditions in the years 2001 and 2002. Of the 801 ESTs, 238 showed significantly similar patterns of expression in at least one treatment: 20 in the elevated CO2 treatment between year-replicates, 167 in the elevated O3 treatment and 111 in the combined elevated CO2 and O3 treatment. We then calculated the mean expression ratio of the 238 selected ESTs in each treatment and flagged the genes that showed qualitatively different expression patterns between year-replicates (Table S1). From among the ESTs with significant levels of expression, we distinguished those that showed a level of regulation of at least 1.5-fold and selected 185 ESTs that were up- or down-regulated in at least one of the treatments: three (two up- and one down-regulated) in the case of the elevated CO2 treatment; 95 (88 up- and seven down-regulated) in the case of the elevated O3 treatment and 51 (44 up- and seven down-regulated) in the case of the combined elevated CO2 and O3 treatment. The relatively small number of genes showing a significant regulation profile between 2001 and 2002 in the elevated CO2 treatment highlights once again the high variability in this treatment and the differences affecting expression in leaves between the two years, when compared to the numbers of significantly regulated genes in the two other treatments. The Cyber-T procedure allowed identification of false positives within calculated P-values. This calculation was not possible for the CO2 conditions because of the variability in the data distribution between the 2001 and 2002 year-replicates (see PPDE column in Tables S2, S3 and S4). For the O3 and combined CO2 and O3 treatments, only less than 1% of the genes, in one case as in the other, were regulated by a 2-fold level or more (Table 2). If regulation levels did not exceed 3-fold in the mean ratios calculated, the highest regulation levels measured for 2001 or 2002, considered separately, reached 12-fold (O3 treatment in 2001). Another important observation here is the small number of genes down-regulated in the three treatments compared with control conditions (less than 5% of significant genes in each treatment).

Table 2.  Regulation of gene expression for selected significantly regulated genes in leaves of aspen (Populus tremuloides) grown under elevated CO2, elevated O3 and combined elevated CO2 and O3 concentrations compared with control conditions (C)
Clone IDAccession no.Gene functionCO2/Cln P-valueO3/Cln P-valueO3 +  CO2/Cln P-valueSOM clusterRNA blot
  1. Associated ln P-values < 0.05 as calculated by the Cyber-T program between year-replicates 2001 and 2002 are given along with mean ratios. These genes were selected from Table S1 for their high level of regulation of expression. Rows are sorted by level of regulation measured in the three different treatments. Signal ratios < 1.0 were inverted and multiplied by −1 to aid their interpretation. Mean ratios highlighted in gray indicate genes that showed similar expression patterns in years 2001 and 2002. The self-organizing map (SOM) cluster column indicates the corresponding cluster in Fig. 2. The asterisk indicates genes tested by RNA blots.

  2. ACC oxidase, 1-aminocyclopropane-1-carboxylate oxidase.

F02G09CA820757Hypothetical protein 1.760.01229 1.59 1.460.03878c12 
R18B06CA822984Cysteine protease 1.370.01332 1.880.049041.790.03431c12*
R75F07CA826280ACC oxidase−1.220.00952 1.270.029431.86 c11*
RA01F02CA821251Ribulose-5-phosphate-3-epimerase−1.430.04338 1.07 −1.08 c4 
R74G09CA826232Aquaporin−1.580.0238 1.07 1.02 c8 
R10A04CA822570Wound-induced protein 1.38  2.570.03381.89 c13*
R04C10CA822148Calmodulin 1 1.29  1.940.048081.620.00365c12 
R12E11CA822712Sucrose synthase 1.22  1.910.045471.62 c12 
R27H05CA823548Ubiquitin/ribosomal protein S27a 1.14  1.870.025791.74 c12 
R08B12CA822430Zinc-finger protein 1.27  1.860.007121.620.02152c12 
R33H02CA823880Histone H4 1.56  1.840.0262.07 c13 
R06C05CA822298Histone H3 1.41  1.840.002951.95 c13 
R22D07CA823207Ribosomal protein L8 1.17  1.820.007621.560.0069c9 
R25F07CA823402Inorganic pyrophosphatase 1.21  1.810.028281.390.04062c9 
R21B12CA823125ATP-citrate lyase 1.37  1.810.00331.660.03525c12 
F11C02CA82116360S acidic ribosomal protein L12 1.09  1.80.038421.430.01225c9 
R10G01CA822614Kinetochore-associated protein Skp1−1.03  1.80.016351.360.0438c8 
R03D06CA822075Alcohol dehydrogenase−1.01  1.780.026341.31 c8 
R02H08CA822034Glutamine synthetase−1.07  1.690.041071.4 c8 
F11A19CA821143Rubisco small subunit−1.27 −1.980.00052−1.470.03338c0*
R71C08CA826021Aquaporin−1.18  1.55 2.460.0153c11 
R77F06CA826403Formate dehydrogenase−1.12  1.99 2.360.04624c11 
R75C04CA826255G3PDH−1.170.02355 1.490.013272.340.00628c11 
R46C01CA824657POP3 wound-induced protein 1.07  1.680.00782.220.0003c13*
R57F07CA825362Naringenin-chalcone synthase 1.02  1.47 2.110.04185c13 
R57C09CA825341Mitochondrial phosphate transporter−1.31  1.43 2.050.04031c11 
R73E09CA826154Dehydrin−1.04  1.45 1.970.0112c11 
R20E04CA823085Beta tubulin 1 1.3  1.650.049731.910.02682c13 
R20A02CA823050Elongation factor eF-2 1.07  1.48 1.910.04272c10 
F11B12CA821155Thaumatin-like protein 1.63  1.63 1.870.03375c12 
R17C05CA822967Hypothetical protein−1.02  1.720.025181.770.01793c11 
R25G01CA823406S-adenosylmethionine synthetase 1.2  1.35 1.770.03061c10*
R16G02CA822945Alpha-6 tubulin 1.33  1.630.016471.750.03363c12 
R16G03CA822946Xyloglucan endotransglycosylase 1.78  1.440.013271.750.0078c12 
R77F05CA826402Polyubiquitin 1.49  1.99 1.710.04684c12 
F11B16CA821159Carbonate dehydratase 1.02 −1.37 −1.820.03151c1 
F02E05CA820743Rubisco small subunit−1.34 −1.780.00123−1.880.04585c0*

The reduced number of ESTs showing consistent expression in the 2001 and 2002 year-replicates could reflect a true environmental effect of the treatments on aspen clones in the FACE facility. We also observed quantitative differences between the years 2001 and 2002 rather than qualitative changes in regulation of gene expression. This observation may be a result of sampling effects and environmental conditions in the hours or days before sampling. These quantitative differences, which could be caused by another stress (biotic or abiotic) in one year and not the other, complicate the analysis but also strengthen the validation of the significant regulation of gene expression detected for the different treatments in the years 2001 and 2002. For example, there was some infestation by tent caterpillars at the FACE site and this may have had some effect on defense gene expression levels, even in control rings. This biotic stress may also have partly accounted for the small number of significant genes found in some of the treatments (e.g. CO2).

Cluster analysis of gene expression

We grouped the expression profiles of the 238 significantly regulated ESTs in the elevated CO2 and/or O3 treatments (alone or in combination) with the program genecluster 2. This allowed us to recognize features in complex, multidimensional gene expression data and classify them into groups of genes sharing similar patterns of regulation in different treatments (Tamayo et al., 1999). Figure 2 presents the 12 clusters derived by genecluster 2 (using a 4 × 4 grid with 16 seeds and 500 000 iterations), and the full list of genes included in each cluster is given in Table S1. As a consequence of the low regulation observed in the CO2 treatment, most of the clusters corresponded to genes more highly expressed under elevated O3 alone (cluster (c) 4, c6 and c9) or in combination with CO2 (c3, c7, c8, c10, c11, c12 and c13). Of the significantly regulated ESTs, only a few showed an up- or down-regulated profile in elevated CO2, alone or in combination with O3, when compared with the treatment with elevated O3 alone. This last point underlines the substantial impact of elevated O3 alone, even on O3-tolerant aspen clone 216, in the long term. The various SOM profiles could be reduced to a limited set of expression patterns based on the graphical outputs. For example, SOMs c0 and c1 could be grouped, and, as presented in Table S1, these clusters consisted of ESTs encoding similar functions. Occurrences of different cellular functions could be detected in other clusters (e.g. c7, c8 and c11), but most of the clusters shared a similar pattern, which corresponded to higher expression in elevated O3, alone or combined with CO2, than in elevated CO2 alone.

Table 2 presents a subset of the most up- or down-regulated genes (1.5-fold regulation) presented in Table S1 in at least one treatment compared with control conditions in the years 2001 and 2002. Gene expression determined by RNA blot is also indicated in Table 2 and the SOM clusters generated are given in Fig. 2. This subset of genes is representative of the global response observed for the different treatments. Few genes were significantly regulated in the CO2 treatment, while a much larger number of genes were up- or down-regulated in the O3 treatment, alone or in combination with CO2. In the CO2 treatment, up-regulation of genes encoding a cystein protease and a hypothetical protein was observed, while genes encoding 1-aminocyclopropane-1-carboxylic acid (ACC) oxidase, an aquaporine and ribulose-5-phosphate epimerase were down-regulated. As shown in the SOM clustering profiles, most of the functions altered in expression in the O3 treatment were similarly altered in the combined treatment with CO2. Up-regulated functions encoded stress-related proteins such as wound-induced proteins, cytoskeleton elements (alpha and beta tubulin) and the secondary metabolism enzyme naringenin-chalcone synthase. Some important functions for leaf physiology were down-regulated in similar ways in all treatments, such as the rubisco small subunit. Several genes encoding enzymes falling into the same cellular categories as presented in Table 2 were also found in the various treatments, but with lower levels of regulation (< 1.5-fold).

RNA blot analyses

To validate observations obtained using the cDNA array approach we used a traditional molecular method. We selected different genes based on their treatment-specific profiles and tested them by northern blot analysis, which gives a good indication of the transcript abundances in the various treatments (Fig. 3). RNA blot analyses for the rubisco small subunit, cystein protease, POP3 wound-induced protein, Cu/Zn superoxide dismutase, ACC oxidase, S-adenosylmethionine synthetase and wound-induced protein genes correlated with expression levels observed in the cDNA array. Thus, the usefulness of nylon-based cDNA arrays for global gene expression profiling in identifying genes with treatment-specific expression patterns from samples harvested in field conditions was confirmed.

Figure 3.

RNA blot analyses for validation of cDNA arrays. Hybridization of 10 poplar cDNA probes to total RNAs extracted from aspen (Populus tremuloides) grown in control (lane 1), elevated CO2 concentration (lane 2), elevated O3 concentration (lane 3) and combined elevated CO2 and O3 concentration (lane 4) treatments in the year-replicate 2002 is shown. Total RNAs, isolated from leaves of aspen clone 216 grown in the different conditions of the FACE project (Karnosky et al., 2003), were hybridized with selected 32P-labeled cDNA inserts and the 26S rDNA encoding gene. rRNAs on gel are shown as quality control of electrophoresis. ABP, auxin-binding protein; rubisco, ribulose bisphosphate carboxylase/oxygenase small subunit; GST, glutathione S-transferase, POP3, POP3 wound-induced protein; Cu/Zn SOD, copper/zinc superoxide dismutase; ACC oxidase, 1-aminocyclopropane-1-carboxylate oxidase; SAM synthetase, S-adenosylmethionine synthetase; WIP, wound-induced protein.


Ozone causes an oxidative burst in plants, which in turn produces reactive oxygen species (ROS) and induces a series of signaling pathways leading to apoptosis. Under elevated O3, a number of ESTs that are involved in signal transduction, defense, and the cell cycle, particularly in senescence or cell death, were up-regulated in trembling aspen growing in the FACE facility. These ESTs included wound-induced proteins, proteases, ethylene biosynthesis-related genes such as ACC oxidase, water stress-induced proteins and signaling proteins. ACC oxidase oxidizes ACC to ethylene. Ethylene is a signaling molecule that stimulates processes leading cell death, including senescence, in plants. Generally, in elevated O3 environments, high concentrations of ethylene are produced (Wolfgang et al., 2002). Senescence-associated genes (SAGs) that were highly expressed in elevated O3 were cysteine protease (Lohman et al., 1994), glutamine synthetase (Bernhard & Matile, 1994), and chloroplast 30S ribosomal protein S7. The gene encoding the POP3 protein is a wound-inducible gene also related to defense (Van Damme et al., 2002), which was up-regulated under elevated O3. A thaumatin-like protein-encoding gene was also up-regulated under elevated O3. These proteins also fall into the plant defense category and are known as pathogenesis-related proteins (S. Duplessis, unpublished).

One gene that was down-regulated gene under elevated O3 corresponds to auxin-binding protein, which is postulated to be responsible for cell expansion and growth (Jones & Herman, 1993). Many genes involved in photosynthesis were also down-regulated under elevated O3, including genes coding for rubisco activase, the small subunit of ribulose bisphosphate carboxylase oxydase (rubisco), the chlorophyll a/b binding protein, and photosystem II (PSII) oxygen-evolving enhancer protein 3. The down-regulation observed for many genes involved in photosynthesis under elevated O3 may contribute to the reduced photosynthetic rates observed under elevated O3 (Noormets et al., 2001a,b). In addition, we detected evidence of premature senescence being triggered under elevated O3 (Sharma et al., 2003).

It is well known that O3 causes premature foliar senescence and early leaf abscission (Landry & Pell, 1993; Volin & Reich, 1996; Nali et al., 1998). Genes up-regulated under elevated O3 that we believe to play a role in the early senescence response included defense genes, antioxidant genes, genes involved in photorespiration or signal transduction, genes encoding wound-induced proteins, senescence-associated genes, and genes involved in the phenylpropanoid pathway associated with anthocyanin production (Winkel-Shirley, 2003). It is interesting to note that similar functions were regulated in the transcriptome analysis of the model plant Arabidopsis under short-term O3 exposure (Tamaoki et al., 2003), such as plant defense genes (mainly up-regulated) and energy-related genes (down-regulated).

Under elevated CO2, up-regulated genes included photosynthetic genes encoding chloroplast 30S ribosomal protein and PSII and photosystem q(b) proteins, important genes for growth, and genes encoding auxin-binding proteins. This suggests that the photosynthetic machinery continues to produce larger amounts of transcripts in aspen trees growing under long-term exposure to elevated CO2. Increased production of gene transcripts involved in photosynthesis is correlated to increased photosynthetic rates, as seen in physiological studies of clone 216 (Noormets et al., 2001a,b; Sharma et al., 2004). Another EST showing higher expression under elevated CO2 was xyloglucan endotransglycosylase, which is responsible for the wall-loosening required for plant cell expansion (Fry et al., 1992). This expression is correlated with the up-regulation of genes coding for various elements of the cytoskeleton associated with growth, such as the alpha and beta subunits of tubulin, and several actin-depolymerizing factors. The expression of these genes may contribute to larger leaf size under elevated CO2 exposure, as seen in many physiological studies (Ferris et al., 2001). Genes encoding auxin-binding proteins were up-regulated under long-term elevated CO2. The auxin-binding protein is known to mediate cell expansion and possibly the cell cycle involved in plant growth (Timpte, 2001). In many physiological studies, stimulation of growth (Tjoelker et al., 1998b; Isebrands et al., 2001) and increased cell expansion (Taylor et al., 2003) have been reported with elevated CO2 exposure. Increased growth under elevated CO2 could be a result, in part, of triggering of the auxin-binding gene. Genes such as ABP were consistently found to be up-regulated over the years of this study under elevated CO2. However, it is not possible to determine whether differences were attributable to cross-hybridization between different auxin binding protein (ABP) mRNA species or to different regulation profiles within this large gene family.

Down-regulated genes under elevated CO2 included drought-induced aquaporin plasma membrane intrinsic protein PIPa2, which is responsible for water movement across the membranes. The reduced transcript abundances of these ESTs suggest that plants may be able to manage water efficiently in an elevated CO2 environment. Tjoelker et al. (1998a) reported that plants had better water management under elevated CO2, as carbon fixed per unit area in the leaf was greater than the water loss, contributing to increased water-use efficiency.

Expression of the phenylpropanoid pathway gene flavanone 3-beta-hydroxylase, which is involved in flavanoid metabolism, was also significantly reduced. Flavanoids consist of various secondary metabolites such as anthocyanins, which give color to flowers and leaves and are regulated by environmental factors or expressed as defense responses to stresses (Winkel-Shirley, 2003). Other genes coding for various enzymes of the secondary metabolism were down-regulated in the elevated CO2 treatment, such as chalcone-flavonone isomerase and cinnamate-4-hydroxylase. These genes also showed similar patterns of expression in all the treatments and were up-regulated under elevated O3 and combined elevated CO2 and O3. Another secondary metabolism gene, naringenin-chalcone synthase, was overexpressed in these treatments but not under elevated CO2 alone. Rubisco activase, which controls the overall activity of photosynthesis by regulating enzyme rubisco, showed reduced expression under elevated CO2. This may suggest that, under elevated CO2 treatment, the photosynthetic rate can become limiting because of a lack of availability of the rubisco enzyme to fix CO2, as CO2 levels become saturated inside the leaf tissue in an elevated CO2 environment (Noormets et al., 2001a).

The application of pattern-specific SOM clustering allowed us to obtain lists of genes that are treatment specific. This method of calculation finds genes with similar expression patterns (without regard to the magnitude of ratios), where up- and down-regulation ratios are determined between 1 and 0. Using this method, the cDNA array global expression profiles can be streamlined and highly defined and reduced to a few critical genes that are responsible for treatment-specific responses. These genes would serve as good markers for future studies to elucidate the genotypic differences among aspen clones with differing tolerance to elevated CO2 and O3, alone and in combination. Validation by RNA blots confirmed the utility of this approach to identify critical genes responsive to specific treatment effects.

Under combined elevated CO2 and O3, there have been reports of amelioration of the negative O3 response in aspen (Volin & Reich, 1996). However, our results with clone 216 showed that CO2 in some cases exacerbated the responses observed under elevated O3, as shown by gene expression profiles in clusters c3, c10 and c13 (Fig. 2). In this study, the expression of some genes under the combined treatment did not show a reversal of the effects of O3 compared with the individual gases. At elevated O3, genes associated with defense and secondary metabolism pathways were up-regulated under the combined treatment and a number of photosynthesis genes were down-regulated, indicating that elevated CO2 was not able to ameliorate all the effects of elevated O3. While the clustering analysis showed that there were no specific profiles related to the combined treatment of CO2 and O3, it is also interesting to note that, under the combined treatment of CO2 and O3, more genes were up-regulated than in the treatment with O3 or CO2 alone.


Our results suggest that 5-yr-old aspen trees exposed to elevated CO2 and O3, alone and in combination, continue to show substantial differences in gene expression patterns. These patterns fit the overall patterns of aspen physiology and growth as summarized by Karnosky et al. (2003). It was also evident that elevated CO2 was not able to ameliorate all the negative effects of O3. We also found in our study that the interacting effects of elevated CO2 and O3 resulted in regulation of expression of genes that were not seen with the individual gas treatments. Further analyses are required on several aspen clones with varying degrees of ozone tolerance to determine how the other genotypes of aspen respond to both elevated O3 and CO2, and to determine which genotypes are best suited to adapt to the changing global conditions of elevated CO2 and O3.

In this environmental genomic study, we also observed expression profiles for some genes that showed different transcript abundances in RNA sampled in the years 2001 and 2002. These contrasting profiles highlight the importance of the contribution of other abiotic and/or biotic stresses to gene regulation, in addition to the treatments applied in the FACE project. However, the strong correlation with physiological observations also emphasizes that global gene expression analysis is possible with samples from trees growing in field conditions. This approach lays the foundations for future studies on ecological genomics.


We thank Dr A. Kohler for providing the poplar 4608-element nylon cDNA arrays and for valuable discussions during the writing of the manuscript. We also thank the two anonymous referees for their comments and suggestions on the manuscript. Part of this work was sponsored by the Office of Science (BER), U.S. Department of Energy Grant No. DEFG02-04ER63792, to DFK and GKP. The research utilized, in part, the DNA Sequencing and Functional Genomics Facilities at INRA-Nancy financed by the INRA and Région de Lorraine through the Institut Fédérateur de Recherche no. 110.

Supplementary material

The following material is available as supplementary material at

Table S1 List of the 238 significantly regulated ESTs analyzed in our expression analysis

Table S2 Gene expression data for CO2 treatment

Table S3 Gene expression data for O3 treatment

Table S4 Gene expression data for CO2+ O3 treatment