The transcriptome of Populus in elevated CO2

Authors


Author for correspondence: Gail Taylor Tel: +44 23 80592335 Fax: +44 23 80594459 Email: g.taylor@soton.ac.uk

Summary

  • • The consequences of increasing atmospheric carbon dioxide for long-term adaptation of forest ecosystems remain uncertain, with virtually no studies undertaken at the genetic level. A global analysis using cDNA microarrays was conducted following 6 yr exposure of Populus × euramericana (clone I-214) to elevated [CO2] in a FACE (free-air CO2 enrichment) experiment.
  • • Gene expression was sensitive to elevated [CO2] but the response depended on the developmental age of the leaves, and < 50 transcripts differed significantly between different CO2 environments. For young leaves most differentially expressed genes were upregulated in elevated [CO2], while in semimature leaves most were downregulated in elevated [CO2].
  • • For transcripts related only to the small subunit of Rubisco, upregulation in LPI 3 and downregulation in LPI 6 leaves in elevated CO2 was confirmed by anova. Similar patterns of gene expression for young leaves were also confirmed independently across year 3 and year 6 microarray data, and using real-time RT–PCR.
  • • This study provides the first clues to the long-term genetic expression changes that may occur during long-term plant response to elevated CO2.

Introduction

On a global scale, forest biomes account for approximately half of all carbon held in terrestrial vegetation (Read, 2002), thus they remain central to our understanding of C fluxes to and from terrestrial pools, and to the development and maintenance of C in land-based C sinks (Grace, 2004). Recent patterns of C exchange suggest that the development of an increased net C sink, globally, can largely be attributed to the northern latitudes, and that a significant component of this is the result of rising concentrations of atmospheric CO2. In Europe, for example, forests are known currently to be a net C sink, with an estimated C-sequestration rate of 363 Tg C m−2 yr−1 (Falge et al., 2002). The longevity of this sink has yet to be determined. Although effective long-term C accumulation must involve slow-release C sinks found in the soil, C capture by trees initially depends on the display of foliage for light absorption and CO2 uptake in the process of photosynthesis. It is well established, from short-term controlled environment studies and exposure of small trees over several years, that elevated [CO2] is likely to result in a stimulation of tree physiology (Gunderson & Wullschleger, 1994; Taylor et al., 1994; Norby et al., 1999) and consequent increases in tree productivity. However, the limitations of small-scale experiments are well known, and few studies have considered the underlying genetic mechanisms that may determine these responses. There is a pressing need to understand more about long-term adaptation and genetic change in future [CO2], particularly for adaptive traits that are relevant to plant productivity and ecological characteristics that determine survival, fitness and interactions with pests and pathogens (Ward & Kelly, 2004).

Long-term forest ecosystem experiments were developed in the 1990s and are now available to answer some of these questions. The forest FACE experiments use a system of open-air fumigation (Hendrey et al., 1999; Miglietta et al., 2001) enabling a large number of trees to be exposed to increased atmospheric [CO2], with good control, allowing the whole ecosystem to be placed under study (Ellsworth et al., 1995; Wullschleger & Norby, 2001; Karnosky et al., 2003). The strength of the FACE approach is that whole ecosystems may be studied, largely undisturbed during long-term exposure that has the potential to last for several years or even decades. Coupled with this development in long-term forest FACE experiments are developments in genomics. To date, genome studies have been used to discover and classify gene functions but few ‘ecological genomic’ studies have been completed, although microarrays have been used successfully to examine gene expression in plants during, for example, organ development (Zhu et al., 2001; Moseyko et al., 2002); environmental stress (Chen et al., 2002); and in response to hormone induction (Che et al., 2002). Transcript profiling has a valuable role to play in the new science of evolutionary and ecological functional genomics, as it can provide an insight into likely candidate genes that may be involved in the control of traits of ecological relevance (Feder & Mitchell-Olds, 2004). Unsuspected and considered genes may be revealed from the thousands of genes available on an array (Gibson, 2002). No data have been reported previously for tree ecosystems exposed to future CO2 concentrations, largely because of the limited availability of large-scale expressed sequence tag (EST) sequencing projects and the production of microarrays in nonmodel plant species. But Populus is an exception, now recognized as the ‘model tree’ (Wullschleger et al., 2002), with the full genome sequence available (http://genome.jgi-psf.org/Poptr1/Poptr1.home), following the development of a large collection of ESTs (Sterky et al., 1998, 2004; Bhalerao et al., 2003), the production of a cDNA microarray (Andersson et al., 2004), and the availability of segregating and natural populations (Taylor, 2002). Such a resource has never before been employed to examine the detailed responses of field-grown plants to predicted future climates and this is an area where, in future, more research should be forthcoming. As plants are sessile and must adapt to changing environments in order to flourish, it can be anticipated that plant gene expression will alter in response to long-term environmental change, and that this will contribute to altered growth and physiology and, in the long term, will lead to adaptive responses at the level of the gene and altered plant fitness.

Here we report a first glimpse of some of the genes that may be most sensitive to [CO2] supply, following a genomic analysis undertaken in one of the forest FACE experiments (POPFACE/EUROFACE) using Populus × euramericana, during years 3 and 6 of exposure to elevated [CO2].

Materials and Methods

Experimental site and growth conditions

The POPFACE/EUROFACE site, located in central Italy, is a 9 ha 3-yr-rotation coppice plantation of fast-growing poplar (Populus spp.), planted during spring 1999. Three of six 350 m2 experimental plots within the plantation are enriched with pure CO2 providing a target concentration of 550 ppm from April to October each year, described by Miglietta et al. (2001). Throughout the fumigation [CO2] measured at 1 min intervals has been within 20% of the target at least 80% of the time. Throughout the experiment, the entire area and all plants have been drip-irrigated. Irrigation has been monitored to ensure equal application in each experimental plot and applied to match transpiration, between 6 and 10 mm d−1.

Each plot contains three species of poplar: in year 1 P. ×euramericana (P. deltoides × P. nigra, clone I-214) was shown to be the most sensitive to C supply (Ferris et al., 2001) and became the target for detailed, genome-wide analysis in year 3 (summer sample, August 2001, data available in Appendices S1 and S2, available online as supplementary material) and year 6 (spring and summer samples, May and July/August 2004, respectively, data available in UPSC-BASE, see below for details). Here we present a detailed analysis for year 6 (summer) samples, with some reference to year 6 (spring) and year 3 (summer), for independent confirmation of some of the findings. All data except those of Fig. 4 relate to the year 6 (summer) analysis, with other data available to the reader. In P. × euramericana, during year 1 the canopy was open and there were no leaves with shade-adapted morphological characteristics. During year 2 canopy closure was achieved but mainstem leaves (the leaves in this study) were not shaded (Gielen et al., 2001). During year 3 the canopy was completely closed (Gielen et al., 2003), although mainstem leaves were again not shaded. This period was characterized by strong competition among trees, and mean canopy height reached 8.5 m (Gielen et al., 2003). During year 4 postcoppicing, branch and foliage cover was dense at all levels in the canopy. During year 5 the canopy was completely closed, although mainstem leaves were not shaded and foliage was less dense lower in the canopy than in year 4. During year 6 the canopy was closed, although mainstem leaves were not shaded and conditions were similar to year 3.

Figure 4.

Independent confirmation of microarray results from year 3 (summer) and year 6 (spring, summer), microarray data (closed bars) for Rubisco (PU11281) and expressed from real-time RT–PCR (open bars).

Sampling of leaves at the POPFACE site

Leaf age was carefully defined by measuring leaf plastochron index (LPI: Taylor et al., 2003), and leaf LPI 3 (young) and LPI 6 (semimature) were sampled on 22 August 2001 (year 3) and 20 May, 20 July, 6 August and 31 August 2004 (year 6) between 15.00 and 18.00 GMT, with sampling confined to within 2 h on a single day, when ambient temperature varied by < 2°C (e.g. 21.6 and 19.7°C on 2 August 2001). Sampling days were chosen as clear and cloudless on each occasion, and daily records of temperature and irradiance confirmed this (data not shown). For each sample point, all leaves were sampled on a single day, often using several researchers simultaneously sampling different trees from scaffolds giving direct canopy access. For each CO2 treatment (ambient and elevated) and leaf age, between four and six biological replicates (trees) were sampled in each treatment plot, giving a total biological replication of 12–18 leaves sampled from individual trees, per treatment, of carefully defined developmental age. RNA was extracted from each leaf separately, as described below, giving > 200 biological replicates across the 2 yr of sampling. Leaves were immediately flash-frozen in liquid N and stored at −80°C.

RNA preparation

RNA was prepared from individual leaves according to Chang et al. (1993) with the following modifications: no spermidine was used in the extraction buffer, and 2.67%β-mercaptoethanol was used. An additional extraction step was performed after precipitation with 2.5 m LiCl. RNA concentrations were determined spectrophotometrically using a NanoDrop ND100 (NanoDrop Technologies, Delaware, USA) and RNA quality was assessed using a bioAnalyser (Agilent 2100 Bioanalyzer, Agilent Technologies, Waldbronn, Germany). Biological replicates (leaves, as described above) were extracted individually, but RNA preparations were pooled for the second round of experiments (using the July/August samples) before analysis. The pools consisted of 7.1 µg total RNA from each of 17 replicates of LPI 3 in ambient [CO2]; 5.2 µg total RNA from each of 23 replicates of LPI 3 in elevated [CO2]; 7.1 µg total RNA from each of 17 replicates of LPI 6 in ambient [CO2]; and 8 µg total RNA from each of 15 replicates of LPI 6 in elevated [CO2].

cDNA synthesis, labelling and purification

The experiment was performed using the standardized microarray-analysis pipeline deployed at the Umeå Plant Science Centre (UPSC) microarray facility (A.S. and coworkers, unpublished). cDNA was synthesized separately from total RNA in each sample/pool. Total RNA (30 µg) suspended in 10 µl diethylpyrocarbonate (DEPC)-H2O was denatured with 1 µl Oligo(dT)-anchor (Invitrogen, Paisley, UK) at 65°C for 5 min, and cooled on ice. mRNA was then reverse-transcribed with 6 µl 5 × RT buffer, 1 µl 50 × dNTP mix (25 mm dA-, dC-, dGTP, 20 mm aa-dUTP, 5 mm dTTP), 3 µl 10 mm DTT, 1 µl RNAse inhibitor (30 U, Invitrogen), 1.5 µl Superscript II (300 U, Invitrogen) and 7.5 µl DEPC-H2O by incubating at 42°C for 3 h. The reaction was stopped with 10 µl 0.5 m EDTA, RNA degraded by adding 10 µl 1 m NaOH and incubating at 65°C for 15 min. The reaction was then neutralized by addition of 50 µl 1 m Hepes pH 7.0. cDNA was purified using Microcon columns according to the manufacturer's specification (Y30 columns, Millipore, MA, USA). cDNA was collected from the membrane and dried in a speedvac (DNA SpeedVac, ThermoSavant, Boston, MA, USA). cDNA pools were created for each experimental condition: ambient LPI 3; elevated LPI 3; ambient LPI 6; elevated LPI 6. Dyes were coupled by first resuspending the Cy3/Cy5 (General Electrics Health Care, Uppsala, Sweden) in 90 µl 0.1 m NaHCO3 pH 9.0. Cy3/Cy5 (15 µl) was then added to the dry cDNA. Dyes were coupled in the dark for 2.5 h at room temperature. Cy3- and Cy5-labelled cDNA was purified using a GFX column according to the protocol supplied (General Electrics Health Care). Dyes were eluted into the same collection tube using 2 × 35 µl buffer EB per dye. The eluted volume was reduced to 82 µl in a Speedvac.

Experimental design of hybridizations

A direct comparison experiment between ambient [CO2] and elevated [CO2] samples within LPI 3 and 6 was conducted for year 3 (summer) and year 6 samples, where six randomly chosen biological replicates for each CO2 treatment and leaf age were used with dye swaps. The overview of this experimental design is illustrated in Fig. 1a. A loop design (Vinciotti et al., 2005) was used for the August samples, the main analysis presented here, where direct comparisons were made between LPI 3 and 6 leaves in ambient and elevated [CO2] (top and bottom of the loop) and between LPI 3 leaves in ambient and elevated [CO2] and LPI 6 leaves in elevated [CO2] (left- and right-hand sides of the loop, Fig. 1b). By using linear models with contrast matrices indirect comparisons were possible, as described in detail by Diaz et al. (2003).

Figure 1.

Design of the cDNA microarray experiments. (a) cDNA from leaves of Populus × euramericana trees in their third year of exposure to either elevated (E, 550 ppm) or ambient (A, ≈ 360 ppm) CO2 concentration at two developmental stages, young (3) and semimature (6), at the POPFACE/EUROFACE site was hybridized on a total of 24 two-colour arrays. Each circle represents cDNA synthesized from an independent biological replicate. Direct comparisons were made between samples of the same leaf age (3 or 6) from elevated and ambient [CO2]. (b) cDNA from leaves of P. × euramericana trees in their sixth year of exposure to either elevated or ambient [CO2] at two developmental stages, young (3) and semimature (6), at the POPFACE/EUROFACE site was hybridized in a loop design. cDNA was extracted independently from a number of biological replicates from within each condition, and these cDNAs were pooled to form one pool per condition. Each circle represents a pool of cDNA from all biological replicates from the relevant condition, with the number of biological replicates represented in each pool shown in brackets for each condition. The four conditions (elevated [CO2] young leaf, E3; elevated [CO2] semimature leaf, E6; ambient [CO2] young leaf, A3; ambient [CO2] semimature leaf, A6) were hybridized against each other on four two-colour arrays. The loop was then repeated with dyes in the opposite orientation, resulting in eight arrays with four measurements for each pool. Linear models were then used to calculate in silico comparisons for age and treatment effects.

Microarray design and hybridization

The POP1 and POP2 Populus glass-spotted cDNA microarrays were used for this study. All year 6 hybridizations used POP2, while year 3 hybridizations utilized POP1. The POP1 array contained 13 488 probes representing 35 000 ESTs (Andersson et al., 2004). The POP2 array contains 24 735 probes representing > 100 000 ESTs from 18 tissues (described by Sterky et al., 2004). Array details, sequence details and GenBank accession numbers for all ESTs are available from the UPSC public database (http://www.populus.db.umu.se).

Hybridizations reported here, with the exception of year 3 (see Appendix S2), were performed in an Automated Slide Processor (ASP) (Lucidea ASP Hybridization Station, Amersham–Pharmacia Biotech, Uppsala, Sweden). Prehybridization buffer was 50% formamide, 5 × SSC and 2.5 × Denhardt's solution. The hybridization solution contained the labelled cDNA, 25% formamide, 5 × SSC, 0.22% SDS, 1 µl tRNA and 0.42 µg Oligo-dA(80mer). Wash buffer 1 was 0.8 × SSC, 0.03% SDS; wash buffer 2 was 0.2 × SSC; wash buffer 3 was 0.05 × SSC, 2 mm KPO4. Isopropanol (100%) was used to clean slides after washing. The ASP used a custom-washing script, treating the slides one by one, giving slides an identical wash between hybridization chambers.

Scanning

Arrays were scanned at four settings of increasing laser power and photon multiplier tube (PMT) setting (laser power, 60, 80, 100, 100%; PMT, 70, 70, 70, 80%) at 10 µm resolution, using a ScanarrayLite Microarray Analysis System scanner (PackardBell, Wijchen, the Netherlands). Regression analysis was then applied to the scans to produce a unified data file from all four scans. This increases the dynamic range of intensities for which spot-intensity data can be extracted (Dudley et al., 2002).

Image analysis

Spot data were extracted using genepix (ver. 5.0 Pro: Axon Instruments Inc., CA, USA). Settings for the spot diameter resize feature were set to < 75 and > 150%, and composite pixel intensity (CPI) was set to 300. Data output from genepix was imported into UPSC-BASE (http://www.upscbase.db.umu.se), where this data set can be publicly accessed and downloaded. The May samples (spring, year 6) are stored as experiment UMA-0002; the July/August samples (summer, year 6) are stored as UMA-0035. Both data sets were print-tip loess normalized using scripts integral to UPSC-BASE (A.S. and coworkers, unpublished).

Data analysis

Linear models with B statistics implemented in the LIMMA (Smyth, 2004: http://bioinf.wehi.edu.au/limma/) package for the statistical software r (http://www.r-project.org) were used to identify genes that may be differentially expressed (Diaz et al., 2003). Lists of all genes with positive B values were examined and analysed further. The B statistic indicates the probability of a gene being differentially expressed, with a B value of 0 representing a 50 : 50 chance of differential expression. Here we examined all genes with values > 0. FDR-adjusted P values were calculated to provide additional support to the interpretation of B values, and are also presented.

Principal component analysis

Principal component analysis (PCA) was used to examine the separation of treatment and age effects on the transformed data (Wissel et al., 2003). PCA analysis was performed using simca p (ver. 10: Umetrics, Umeå, Sweden).

Confirmation of microarray results using real-time RT–PCR

Real-time RT–PCR was performed on individual RNA samples previously used to make the RNA pools hybridized on the microarray (July/August, year 6). In total, four randomly selected biological replicates (leaf samples) from each of the four conditions (LPI 3 ambient [CO2]; LPI 3 elevated [CO2]; LPI 6 ambient [CO2]; LPI 6 elevated [CO2]) were used from leaves sampled on 6 August 2004. Four technical replicates were produced per biological replicate.

cDNA synthesis  Before cDNA synthesis, contaminating genomic DNA was digested using the TURBO DNase-free kit (Ambion, Austin, TX, USA) according to the manufacturer's protocol. The cDNA was synthesized following a similar protocol as that for the microarrays. However, the RNA was incubated with oligo(dT)20 primer at 70°C for 5 min, after which only 1 µl SuperScript III (200 U, Invitrogen) and a standard dNTP mix was used for reverse transcription. The initial incubation was 1 h at 50°C, and RNA was degraded using 3 µl 5 m NaOH. The cDNA was cleaned using the QIAquick PCR Purification Kit (Qiagen, Crawley, UK) and the concentration of cDNA was measured on a spectrophotometer (ND-1000 Spectrophotometer, V3.0.0, Nanodrop, DE, USA). Only samples between 2 and 10 ng µl−1 were used for RT–PCR.

Primers  Primers were designed using the programme beacon designer 4.0. From the results of the microarray experiment, primers were designed for the small subunit of Rubisco transcript. The control gene used was a ribosomal protein (PU00602). The primer used for RT–PCR was Rubisco small subunit (PU11281) forward 5′-ATCTCACAGAGCAGGAATTGG-3′; reverse 5′-AGTAGCGTCCATCATAGTACC-3′.

RT–PCR reaction  Each 20 µl reaction included 10 µl SYBR green (Finnzymes, GRI, Braintree, UK), 8.4 µl DEPC-treated water, 0.6 µl primer mix containing 10 µm forward and reverse primers, and 1 µl cDNA template, all of which were pipetted on ice. The plates were run on the DNA Engine Opticon 2 System (MJ Research, Braintree, UK). An initial heating step of 95°C was performed for 10 min at the beginning of each run. The plates were then run at 94°C for 10 s for the denaturing step, annealed at 56°C (the optimum temperature for the primers) for 20 s, and extended at 72°C for 20 s. A melting curve produced at the end of each run was used to check for secondary products (Ririe et al., 1997). The programme linregpcr was used to calculate the PCR efficiencies. Real time RT–PCR was compared with expression from the microarrays, and for this purpose both year 6 sampling times (May and July/August) and year 3 (August) transcripts were compared for PU11281 using expression ratios for both leaf ages.

Results

In the POPFACE experiment, trees exposed to elevated CO2 have been followed for several years. We performed transcriptomics analyses of leaves sampled in years 3 and 6 of the experiment, in July (summer, year 3) and May and July/August (spring and summer, year 6) when the canopy was fully closed. For year 3 and year 6 (spring) samples, individual samples were treated separately; the experimental design is illustrated in Fig. 1a. There is an ongoing discussion in the microarray analysis field as to whether the best practice for this type of array experiment is pooling of many biological replicates before array analysis, or merging of data after the analysis (Shih et al., 2004); here we have compared both approaches. For the July samples we used the second strategy, pooling of RNA preparations from 15–23 biological replicates and running a loop (Fig. 1b). This was done to obtain a comparison between the two designs; also, as a preliminary analysis of the year 6 May (spring) samples indicated that developmental changes may contribute to the transcriptome differences, this design allowed for improved comparisons both between leaf ages (LPI) and treatments. We analysed the results from both designs in parallel with similar findings, showing that the design of the array experiment was not important for the general conclusions. For clarity, we discuss here only the data set generated from the loop experiment. The results from all individual arrays, for the year 3 preliminary analysis, are provided in Appendix S2. The PCA undertaken for the complete data set using year 6 July (summer) samples is illustrated in Fig. 2. This shows that the first component explaining most variability was leaf developmental stage (LPI 3 vs LPI 6), and that the effects of elevated [CO2] were of only secondary importance. There may also be a developmental shift in elevated [CO2] leaves which, at both leaf ages, are shifted to the right of the plot. Tables 1 and 2 confirm these findings, as they detail genes with a high probability of differential expression in response to CO2 treatment for LPI 3 and LPI 6 leaves. Correct statistical analysis of microarray data is critical if type 1 and type 2 errors are to be avoided. Here, as the changes in gene expression in response to [CO2] were small, the matter of statistical treatment is not simple. We have chosen to examine genes with a high probability of differential expression. Given this approach, only eight transcripts were identified in young leaves (Table 1), including those related to cell-wall and membrane biosynthesis and function and calcium signalling. These represent candidate transcripts for further study and confirmation. For semimature leaves (Table 2), 28 transcripts were identified. Importantly, all except seven showed a downregulation in elevated relative to ambient [CO2]. Of note was a transcript for pyruvate kinase, a key regulator of the step between C metabolism and protein synthesis and a number of transcription factors. Two transcripts related to chloroplast biosynthesis and function were also downregulated, as was a heat-shock protein transcript.

Figure 2.

Results of principal components analysis of normalized transcript intensities for ≈ 25 000 cDNA probes, showing that the principal component explaining the majority of the observed variation was leaf age (young vs semimature) and the second component was [CO2] (elevated vs ambient).

Table 1.  Expressed sequence tags (ESTs) with positive B values for LPI 3 (young) CO2 response and associated P statistics
IdentifierAnnotationMBP
  1. All were statistically significant; a positive M value suggests upregulated gene expression in elevated relative to ambient CO2.

PU28532Leucine-rich-repeat family protein   1.622.22260.341
PU27165GDSL-motif lipase/hydrolase family protein   1.891.94740.341
PU20530Endoxyloglucan transferase   1.641.05050.467
PU20437Calcium-dependent protein kinase isoform 2 (CPK2)   2.030.79970.467
PU09556Harpin-induced family protein   1.460.56370.527
PU09305GDSL-motif lipase/hydrolase family protein   2.570.40840.552
PU08476Proline-rich family protein−3.60.36230.467
PU05763Polcalcin, putative/calcium-binding pollen allergen   1.780.04850.564
Table 2.  Expressed sequence tags (ESTs) with positive B values for LPI 6 (semimature) CO2 response and associated P statistics
IdentifierAnnotationMBP
  1. All were statistically significant; a negative M value suggests downregulated gene expression in elevated relative to ambient CO2.

PU10409No annotation available−2.473.4180.0525
PU12448Ras-related GTP-binding protein, putative−5.763.2390.0247
PU28637Potassium transporter (KUP1)−2.403.0190.0591
PU11724Suppressor protein SRP40. Saccharomyces cerevisiae (baker's yeast). SRP40 or YKR092C or YKR412A−3.562.1890.0887
PU00177GATA transcription factor 1 (GATA-1)−1.541.6690.1351
PU08917Chloroplastic RNA-binding protein P67, putative nearly−10.201.3840.0256
PU22860Chloroplast inner envelope protein-related−1.621.0870.162
PU25517Heavy metal-associated domain-containing protein−5.820.9550.0869
PU06984Pyruvate kinase, putative−6.310.9350.0869
PU21852Ras-related protein (ARA-3)/small GTP-binding protein, putative−3.840.7280.162
PU21030Expressed protein1.870.7250.162
PU12050Heat-shock protein, putative strong similarity to heat-shock protein [Arabidopsis thaliana] GI:1906830−1.640.6140.162
PU05072Elongation factor 2, putative/EF-2, putative−4.970.6120.1381
PU06234Adenylate kinase family protein5.250.6010.1381
PU09638T-complex protein 1 alpha subunit/TCP-1-alpha/chaperonin (CCT1)−1.940.5910.162
PU08927Eukaryotic translation initiation factor 3 subunit 10/eIF-3 theta/eIF3a (TIF3A1)−2.260.4940.162
PU22579Nodulin MtN3 family protein−1.350.3800.162
PU10293Endomembrane protein 70, putative TM4 family4.730.3650.162
PU11079COP9 signalosome complex subunit 7ii/CSN complex subunit 7ii (CSN7) (COP15)/FUSCA protein (FUS5) FUSCA5, CSN7, COP153.240.3380.162
PU04471ATP-dependent Clp protease proteolytic subunit (ClpRI) (nClpP5)−4.760.3000.162
PU07311F-box family protein contains Pfam profile: PF00646 F-box domain−1.240.2940.162
PU10127WWE domain-containing protein/ceo protein, putative (CEO) contains Pfam domain, PF02825: WWE domain2.140.2810.162
PU26674CTP synthase, putative/UTP – ammonia ligase, putative−5.460.2120.162
PU2945960S ribosomal protein L11 (RPL11B) ribosomal protein L11, cytosolic, Arabidopsis thaliana, PIR:S490332.200.1910.162
PU0646360S ribosomal protein-related−2.110.1680.162
PU02302Peptidyl-prolyl cistrans isomerase (PIN1)/cyclophilin/rotamase−1.150.1680.162
PU03360Aspartyl-tRNA synthetase, putative/aspartate – tRNA ligase, putative−3.620.1310.162
PU06179No annotation available−1.100.1290.162
PU08744Expressed protein4.000.0820.162
PU08796Expressed protein contains Pfam profile PF04483: protein of unknown function (DUF565)−3.780.0580.162
PU28466ARF GTPase-activating domain-containing protein−1.330.0330.162

A more detailed analysis of multiple ESTs representing the same gene is provided in Fig. 3 (3′ replicates identified from 3′ resequencing) for the small subunit of Rubisco (rbcS), a transcript already known to be sensitive to elevated [CO2]. For these Rubisco transcripts we confirmed this trend of gene expression in young and semimature leaves, respectively, in response to elevated [CO2] by conducting a two-way anova test. In addition, mRNA blotting on RNA prepared from the same material but sampled in year 3 confirmed this trend (data not shown). Independent confirmation of this result is also shown in Fig. 4, where the Rubisco transcripts were assessed from all microarrays and using real-time RT–PCR. For summer leaves from both years 3 and 6, transcripts for Rubisco small subunit in young leaves was upregulated in elevated CO2 according to the microarray data, and this was further confirmed using RT–PCR. For semimature leaves, the positive response to elevated CO2 was always less than in young leaves, and on two occasions a small reduction was observed. Nevertheless, the consistency of this response suggests that these changes are of genuine biological importance, but that the changes documented at the individual gene level are small (less than twofold up or down). Expression profiling using microarrays is good for screening a large number of genes but, as the number of replicates is typically small, it is less likely to detect statistically significant differences for genes that differ only a little in expression level between treatments. It is also apparent from Fig. 4 that some subtle differences in response are likely between spring and summer leaves, perhaps reflecting changes in canopy development with season.

Figure 3.

Fold-change in normalized intensities of multiple expressed sequence tags representing Rubisco small subunit (rbcS) for leaves exposed to elevated (550 ppm) vs ambient [CO2]. anova shows a significant interaction between [CO2] concentration and leaf age.

The significant anova interaction (Fig. 3) was further explored for the whole data set. In Table 3 and Fig. 5 the number of transcripts that were differentially expressed between young and semimature leaves have been classified according to the UPSC–MIPS (Munich Information Centre for Protein Sequence) functional classification. The analysis revealed that developmental stage had a large effect on gene expression in Populus, and that [CO2] had an influence on these changing patterns of gene expression. For example, in ambient [CO2] 13 transcripts within the photosynthetic gene category were upregulated in young relative to semimature leaves, while in elevated [CO2] the number was 16. Further investigation of these interactions between developmental stage and [CO2] is warranted for the categories amino acid metabolism, carbohydrate metabolism and transcription.

Table 3.  Effects of leaf developmental stage (LPI3, young vs LPI6, semimature) on gene expression. The numbers of transcripts showing differential expression (upregulated and downregulated) between LPI3 and LPI6 leaves are shown for elevated and ambient [CO2]. Probability of differential expression was taken from the B statistic as described in the text.
UPSC-MIPS categoryElevated CO2Ambient CO2
updownupdown
amino acid metabolism18115
C-compound and carbohydrate metabolism144152
cell aging0201
cell cycle1532
CELL CYCLE AND DNA PROCESSING1012
cell differentiation0101
CELLULAR TRANSPORT AND MECHANISMS0302
CLASSIFICATION NOT YET CLEAR-CUT5365
Disease, virulence and defense1121
DNA processing17015
electron transport and membrane-associated energy conservation3020
ENERGY0100
fermentation1100
intracellular signalling3033
ionic homeostasis1100
lipid, fatty-acid and isoprenoid metabolism1241
METABOLISM3434
metabolism of vitamins, cofactors, and prosthetic groups2011
mRNA transcription3341
Nitrogen and sulfur metabolism0011
nucleotide metabolism0012
nucleus3101
phosphate metabolism2411
photosynthesis163131
protein modification2020
Protein synthesis2211
protein targeting, sorting and translocation0032
respiration1010
ribosome biogenesis1203
secondary metabolism2221
stress response0310
TRANSCRIPTION0421
translation1010
Transposable elements, viral and plasmid proteins0001
TRANSPORT FACILITATION0000
TRANSPOSABLE ELEMENTS0101
UNCLASSIFIED PROTEINS17353231
Figure 5.

Venn diagram of differently expressed transcripts between young (LPI 3) and semimature (LPI 6) leaves in elevated (550 ppm) and ambient [CO2] for leaves of Populus ×euramericana at the POPFACE/EUROFACE experimental site, revealed by cDNA microarrays. The analysis reveals a set of 93 transcripts that were differentially expressed between leaf ages in both ambient and elevated CO2 treatments.

Discussion

This study provides an initial insight into how gene expression is likely to be altered following long-term exposure of a forest ecosystem to the concentration of atmospheric CO2 predicted for 2050. Populus × euramericana tree growth continued to be stimulated positively during this 6 yr experiment, and this was associated with increased leaf area development (Taylor et al., 2003; Tricker et al., 2004); leaf photosynthesis (Bernacchi et al., 2003; Calfapietra et al., 2003); and enhanced partitioning of biomass to roots (Lukac et al., 2003). It has been suggested that the responses of ‘closed-canopy’ forests to elevated [CO2], such as the one studied here, may be limited as trees become acclimated physiologically, particularly with respect to photosynthesis and the amount and activity of Rubisco (Rogers & Humphries, 2000). However, only limited evidence for this type of acclimation has been found in the POPFACE/EUROFACE experiment, although the magnitude of response to elevated [CO2] has declined somewhat (Bernacchi et al., 2003). Our aim was to use transcript profiling to deduce patterns of gene expression and candidate genes underlying traits of physiological and ecological relevance, some likely to be predicted from previous physiological and biochemical studies, and others highly novel. In P. × euramericana, a fast-growing, indeterminate tree, such traits include altered leaf morphology (size, shape and quality); canopy architecture and longevity (branch formation and growth and proleptic vs sylleptic branching, bud set and flush); and partitioning below ground – all of which may ultimately have a role in determining plant fitness (A. M. Rae and coworkers, unpublished).

Perhaps the most striking finding of this study was that rather few transcripts were significantly affected by the CO2 treatment. Here we focus on the detailed analysis of pooled July/August year 6 samples, the most reliable of the three data sets available from this long-term field experiment. PCA (Fig. 1) revealed that leaf developmental stage was of greater significance than [CO2] in determining the gene-expression response, accounting for the greatest proportion of variation in the data. Other researchers studying photosynthetic gene expression in response to elevated [CO2], using more traditional single-transcript approaches, have found downregulation of the primary CO2 fixation enzyme, Rubisco, at the mRNA level (Makino et al., 2000). This downregulation is considered to occur as a consequence of increased hexose and sucrose cycling acting as sugar signals, as described by Long et al. (2004), so it is clear that this pathway is sensitive to CO2 supply. In our experiment we found the same downregulation in semimature leaves; in contrast, Rubisco transcripts were accumulating in young leaves. A similar tendency was also seen if all genes in the UPSC–MIPS class ‘photosynthesis’ were considered. This opposite pattern of transcript abundance in young and semimature leaves, and the results from the PCA plot analysis, show that the developmental stage of the leaf had a profound effect on gene expression. This may reflect an early photosynthetic competence being attained in these leaves in response to [CO2], a hypothesis for which we have preliminary data (C. A. Raines, pers. comm.), but also suggests that the pattern of leaf development is likely to be accelerated in elevated [CO2], confirming earlier findings (Taylor et al., 2003). This could have an overriding effect in determining gene-expression responses. A consequence is that LPI in ambient and elevated [CO2] is not necessarily comparable and, considering that most previous research on elevated [CO2] has taken place on leaves of similar age (often first mature leaves), the results may be more influenced by indirect effects of [CO2] on developmental stage rather than direct effects on adaptation to [CO2]per se.

The difference in response to elevated [CO2] with leaf age suggests that stage of development is a critical consideration for the design of sampling for transcript profiling, such as that depicted here, and that there is tight genomic control of leaf development as suggested by other studies (Fleming, 2003). The young leaves in this study were considered to be in transition between sink and source status for photosynthetic C gain in this experiment. In contrast, semimature leaves were considered to be optimal for leaf photosynthesis and where the largest stimulations in leaf growth for trees exposed to elevated [CO2] were apparent (Taylor et al., 2003). These developmental differences had a dramatic effect in determining patterns of gene expression, and the loop design employed here allowed us to analyse the effect of developmental stage separately from that of [CO2], revealing a large number of transcripts that differed in expression between young and semimature leaves, and their functional classifications in both [CO2] treatments (Table 3; Appendices S1 and S2). Transcripts in the categories photosynthesis, carbohydrate metabolism, amino acid metabolism, cell cycle and protein targeting were clearly changed with leaf age, and this age effect also differed between ambient and elevated [CO2].

There is strong evidence that, in general, C status of the plant has a large influence on gene expression – plants exhibit some plasticity in gene expression and can respond to CO2. Alterations in plant carbohydrate status differentially regulate photosynthesis pathways (Koch, 1996; Moore et al., 1999), biosynthetic activity (Yu, 1999), growth and organogenesis (Pien et al., 2001), and appear to be involved as coregulators of responses to biotic and abiotic stress (Roitsch, 1999). In particular, sugars regulate gene expression and activate signal-transduction pathways (Sheen et al., 1999; Smeekens, 2000). Seneweera et al. (1995) showed that increased leaf growth in elevated [CO2] in rice was correlated with an increasing content of fructose and glucose, and a greater activity of the sugar-metabolizing enzyme sucrose-phosphate synthase. However, this increased leaf growth was dependent on the plant's ontogeny and the source–sink status of the leaf.

Given these findings, it seems unlikely that plant gene expression following long-term growth in elevated CO2 is ‘acclimated’ with the control of physiology and biochemistry – for example the biochemistry of photosynthesis – occurring somewhere other than at the level of gene expression. It appears likely that there is no specific sensing system for [CO2], but rather that the changes brought about in leaves during long-term exposure to moderately elevated [CO2] result in a modification of the existing leaf developmental pattern. Leaf development is a very plastic process, particularly in poplar, responsive to light, water and temperature. Atmospheric [CO2] appears to modify the leaf developmental pattern within normal limits, but this modification apparently resulted in changes in the physiological responses of the leaf. This plastic response, and associated changes in gene expression, may represent an identification of the most sensitive genes regulating long-term adaptation and evolution. Our next step will be to determine natural variation in these candidate genes, and to consider their likely potential in determining evolutionary change.

In addition to genes involved in photosynthesis, transcripts related to chloroplast biogenesis and function were also downregulated in semimature leaves in elevated [CO2] (Table 2). Drake et al. (1997) illustrated that a loss in Rubisco of up to 40% was possible in elevated [CO2] without any detrimental effect on photosynthetic [CO2] fixation; here at the POPFACE site we see how reduced Rubisco transcript abundance had limited effects on the measured rate of photosynthesis (Bernacchi et al., 2003; Calfapietra et al., 2003).

The analysis used here has identified a small set of transcripts that may be differentially expressed in elevated [CO2], but has also shown that these are distinct in young (LPI 3) and semimature (LPI 6) leaves (Tables 1 and 2). In young leaves, transcripts involved in developmental response, in particular cell-wall loosening and synthesis through the action of xyloglucan endotransglycosylase, may also be upregulated, confirming early findings that growth promotion in elevated [CO2] involves alterations in cell-wall biochemistry and function (Taylor et al., 1994). Calcium-signalling transcripts were also upregulated, and this is perhaps more interesting. Upregulation of the calcium-dependent protein kinase CPK2 in elevated [CO2] may provide a clue to the mechanism through which C metabolism may be altered in elevated [CO2], and also suggests that this may be an interesting gene for further analysis, as it is already known that the activity of the sucrose-cleaving and -synthesizing enzymes sucrose synthase and sucrose-phosphate synthase may be modified by CPKs (Cheng et al., 2002). The details of such control are yet to be elucidated for plants exposed to elevated [CO2]. For semimature leaves, where most of the transcripts showing differential expression were downregulated (Table 2), there were some exceptions that also showed upregulation in elevated [CO2], including transcripts for the enzyme adenylate kinase, an important regulator in determining shifts to glycolysis over gluconeogenesis.

In addition, transcripts identified as differentially expressed between leaf ages were identified that were only present in elevated [CO2], and these were contributing to the interaction, as shown for Rubisco transcripts (Fig. 3). One possibility is that leaf development may be accelerated in elevated [CO2] such that leaves reach maturity and begin to senesce more quickly, and that the patterns of gene expression, particularly for LPI 6 leaves, reflect this. Were the majority of differentially expressed genes downregulated in elevated [CO2] because leaves were more mature? We cannot answer this question yet. However, of genes known to be associated with senescence in Populus (Andersson et al., 2004), only those likely to be involved in downregulation of photosynthesis were influenced in a manner consistent with senescing leaves, and our current research suggests that leaf longevity is actually increased in elevated [CO2]– an effect possibly related to decreased specific leaf area (Tricker et al., 2004).

In conclusion, we have identified developmental and gene-expression changes that occur when trees are exposed to the concentration of atmospheric [CO2] likely in 2050. Few transcripts showed large differences between ambient and elevated [CO2], but many showed a trend of up- and downregulation in young and semimature leaves, respectively. This may be a consequence of accelerated leaf development in elevated [CO2], and this should be considered carefully in the design of future experiments. Nevertheless, we have provided a first look at some of the most responsive transcripts that are altered in elevated [CO2], and these include important regulators of C metabolism and signalling molecules. It appears likely that trees of the future will be genetically distinct from those of present-day [CO2] climates, and we are beginning to understand some of the likely candidates for genetic adaptation.

Acknowledgements

This work was supported by the EC through its Environment R&D programmes within the Fourth and Fifth Framework Programmes for Research as ENV4-CT97-0657 (POPFACE) and EVR1-CT-2002-40027 (EUROFACE), respectively, coordinated by the University of Viterbo and by the Natural Environment Research Council, Department of Environment, Food and Rural Affairs and Hope (grant nos GR9/04077, NFO410 and GO2 to G.T.). P.J.T. and N.R.S. were awarded research studentships from the Natural Environment Research Council (nos GT04/99/TS250 and NER/S/A/2001/06361). This study also contributes to the Global Change and Terrestrial Ecosystems elevated CO2 consortium of the International Geosphere–Biosphere Programme. Research in the laboratory of G.T. is also funded by Department of Energy, USA, project POPGENICS, at the POPFACE site.

Ancillary

Advertisement