Comparative transcriptome analysis of toxic metal responses in Arabidopsis thaliana and the Cd2+-hypertolerant facultative metallophyte Arabidopsis halleri


Stephan Clemens. Fax: +49 345 55821409; e-mail:


Toxic effects of both essential and non-essential heavy metals are well documented in plants. Very little is known, however, about their modes of toxicity, about tolerance mechanisms and the signalling cascades involved in mediating transcriptional responses to toxic metal excess. We analysed transcriptome changes upon Cd2+ and Cu2+ exposure in roots of Arabidopsis thaliana and the Cd2+-hypertolerant metallophyte Arabidopsis halleri. Particularly, three categories of genes were identified with the help of this comparative approach: (1) common responses, which might indicate stable and functionally relevant changes conserved across plant species; (2) metallophyte-specific responses as well as transcripts differentially regulated between the two species, representing candidate genes for Cd2+ hypertolerance; and (3) those specifically responsive to Cd2+ and therefore indicative of toxicity mechanisms or potentially involved in signalling cascades. Our data define, for instance, Arabidopsis core responses to Cd2+ and Cu2+. In addition, they suggest that Cd2+ exposure very rapidly results in apparent Zn deficiency, and they show the existence of highly specific Cd2+ responses and distinct signalling cascades. Array results were independently confirmed by real-time quantitative PCR, thereby further validating cross-species transcriptome analysis with oligonucleotide microarrays.


The ions of both essential and non-essential metals such as copper and cadmium, respectively, can be toxic to plants. The mechanistic basis of toxicity is not understood in detail (Rea, Vatamaniuk & Rigden 2004). Cu is a redox-active metal. Under physiological conditions, two different oxidation states occur, Cu(I) and Cu(II). Thus, supraoptimal Cu ion concentrations can elicit the formation of hydroxyl radicals through the Fenton and Haber–Weiss reactions (Halliwell & Gutteridge 1990). Even less is known about the modes of toxicity of Cd2+ once it is taken up into plant cells, most likely via Ca2+, Fe2+ and Zn2+ uptake systems (Clemens et al. 1998; Connolly, Fett & Guerinot 2002; Perfus-Barbeoch et al. 2002). There are several reports documenting oxidative stress following exposure to high Cd2+ concentrations (Sanita di Toppi & Gabbrielli 1999). However, cadmium ions do not alter their oxidation state. Oxidative stress might therefore rather be an indirect effect of a depletion of reduced glutathione (GSH) due to the synthesis of phytochelatins (PCs), GSH-derived metal-binding peptides (De Vos et al. 1992). Another potential cause of toxicity is the high affinity of Cu and Cd2+ ions to functional groups of biological molecules, in particular SH groups, O- and N-containing groups. The binding can render molecules inactive. Cd2+ ions, for instance, are hypothesized to replace metal cofactors such as Zn2+ ions from proteins or compete with Ca2+ for binding to Ca2+-binding proteins (Stohs & Bagchi 1995). However, as to whether particularly sensitive ‘target’ sites exist for Cu2+ or Cd2+ ions, whose inhibition or inactivation would represent the primary event is not known from any biological system.

Cu and Cd toxicity are physiologically relevant for several reasons. Plant roots can be exposed to varying levels of available Cu ions in the soil (Kabata-Pendias & Pendias 2001). Furthermore, during certain developmental processes such as senescence-associated degradation of plastids, Cu ions are released (Himelblau & Amasino 2001). Cd ions occur naturally in soil and are a major environmental pollutant (Pinot et al. 2000). Uptake by plants is the main source of Cd accumulation in food. Cd responses are known to influence accumulation rates and thereby enter into the food chain. Increased intracellular binding, for instance, drives accumulation of Cd (Clemens et al. 1999).

Upon exposure to potentially growth-inhibiting external concentrations of metal ions, plant cells respond with the activation of detoxification pathways. This activation can be posttranslational or, as in most cases, transcriptional. Well documented is the activation of the PC-dependent detoxification pathway. Synthesis of PCs is activated upon binding of metal ions or metal–GSH complexes to the constitutively expressed enzyme (Vatamaniuk et al. 2000; Maier et al. 2003). In addition, an increase in transcript levels has been reported for TaPCS1 in wheat roots (Clemens et al. 1999) and AtPCS1 in Arabidopsis seedlings after Cd2+ treatment (Lee & Korban 2002). A number of processes upstream of PC synthesis are transcriptionally up-regulated under Cd2+ exposure as well. These include sulphate uptake, sulphate assimilation, cysteine biosynthesis (Dominguez-Solis et al. 2001) and GSH biosynthesis (Xiang & Oliver 1998). Very few comprehensive studies of excess metal responses have been published to date. Pre-dominantly for the yeasts Saccharomyces cerevisiae and Schizosaccharomyces pombe, profiling experiments have revealed Cd2+-responsive fractions of the proteome or transcriptome, respectively (Vido et al. 2001; Chen et al. 2003).

Completely unknown is, how the transcriptional responses to elevated metal levels are mediated in plants. Also not understood is, whether the response is direct or indirect, that is, whether there is a direct sensing of the metal status or rather a sensing of the damage occurring, for instance, through the formation of hydroxyl radicals in the Fenton reaction. In addition, the chain of events mediating a transduction of a signal is unknown. A direct sensing of elevated Cd2+ levels appears unlikely, since no physiological functions are known for Cd2+ with the exception of a Cd-dependent carboanhydrase in Thalassiosira weissflogii (Lane et al. 2005). In contrast to that, a direct sensing of the Cu status of a plant cell can be postulated. Cu ions have acquired a number of biological functions after they became available to biological systems with the advent of oxygen in the atmosphere (Frausto da Silva & Williams 2001). They are essential for various redox reactions and serve as cofactors for many pre-dominantly extracytosolic enzymes such as phenol and ascorbate oxidases. Other Cu-dependent proteins in plants include the electron-carrier plastocyanin and the ethylene receptor (Rodriguez et al. 1999). The uptake and distribution of Cu ions needs to be carefully controlled because of the above-mentioned redox activity. For yeast cells, it was shown in recent years that there are practically no free Cu ions present. Following uptake through tightly regulated transporters, Cu ions are bound by chelators and metallochaperones that act as buffers and function in the distribution of Cu ions to target proteins and organelles (O’Halloran & Culotta 2000). The transcriptional regulation of Cu influx and detoxification systems is mediated by MAC1 and ACE1, two transcription factors whose DNA-binding activity is directly modulated by Cu binding (Winge 2002). No such regulatory proteins are known from plants. No molecules active upstream of the transcriptional regulation of a gene upon toxic metal treatment have been identified to date. Jasmonate has been implicated in the response to excess Cu2+ and Cd2+ (Xiang & Oliver 1998).

A number of plants have evolved naturally selected metal hypertolerance and are able to thrive in metal-contaminated soil. Some of these hypertolerant plants are also able to hyperaccumulate certain metals such as Ni, Zn or Cd (Baker & Brooks 1989). These traits and their potential exploitation for the phytoremediation of soil-metal contamination have in recent years stimulated intensive research mostly on two model systems: Thlaspi caerulescens and the facultative metallophyte Arabidopsis halleri (Krämer 2005). One major objective of these studies is to elucidate the as yet poorly understood molecular mechanisms of metal hypertolerance.

We undertook a transcriptome analysis of rapid cadmium and copper responses in Arabidopsis thaliana and A. halleri roots with three major objectives: (1) by comparing the effects of Cu2+ and Cd2+ exposure on transcript levels in roots, we attempted to obtain information about the modes of toxicity of a redox-active and a non-redox-active metal ion; (2) by comparing the responses of a normal plant and a zinc- and cadmium-hyperaccumulating metallophyte such as A. halleri, we aimed at identifying shared and differential responses to metal ions which might lead to a better understanding of metal-tolerance mechanisms; and (3) based on the assumption that a metal-responsiveness of putative signal-transduction components might indicate a functional involvement of the encoded proteins in mediating metal responses, we hoped to identify candidate genes to be analysed further through reverse-genetic approaches. We report on the rapid transcriptional responses of A. thaliana and A. halleri roots to exposure to different Cd2+ concentrations and for comparison to Cu2+. For several genes the microarray results were validated by quantitative real-time PCR.


Plant material, growth conditions and treatments

A. thaliana Col-0 plants were grown in a hydroponic system with 1/10 strength modified Hoagland's medium: 0.28 mm Ca(NO3)2, 0.10 mm (NH4)2HPO4, 0.20 mm MgSO4, 0.48 mm KNO3, 0.03 µm CuSO4, 0.08 µm ZnSO4, 0.50 µm MnCl2, 4.60 µm H3BO3 and 0.01 µm MoO3. Fe was supplied as N,N′-di-(2-hydroxybenzoyl)-ethylenediamine-N,N′-di-acetic acid (Fe-HBED; 5 µm). Seeds were germinated on solid medium-filled PCR tubes, and after 3 weeks, were transferred to 1.5 L pots (three plants per pot) in a growth chamber with a 8-h light/16-h dark cycle, at 22 °C. A. halleri (L.) O’Kane and Al-Shehbaz (formerly Cardaminopsis halleri[L.] Hayek) plants were collected at a site near Langelsheim in the Harz mountains, Germany, which is metal-contaminated due to medieval mining activities. Plants were propagated vegetatively on soil. Cuttings of A. halleri plants were cultivated on the same medium for about 2 weeks. After rooting, plants were transferred to 1.5 L pots (three plants per pot). The plants were grown in a growth chamber with a 16-h light/8-h dark cycle at 22 °C. For both species, the medium was changed every week and 24 h before treatment.

Hydroponic cultures of A. thaliana were treated with 10 µm Cd2+ (= low cadmium), 50 µm Cd2+ (= high cadmium) and 10 µm Cu2+ by transferring the plants to heavy-metal media containing 1/10 strength modified Hoagland's medium. Control plants were transferred to a fresh medium without added metal. After 2 h, the roots were harvested for total RNA extraction. Hydroponic cultures of A. halleri were treated with 25 µm Cd2+ (= low cadmium), 125 µm Cd2+ (= high cadmium) and 10 µm Cu2+ in the same manner.

For metal tolerance assays, seeds of A. thaliana and A. halleri were surface-sterilized (70% ethanol for 2 min, then 10 min in 10% sodium hypochlorite) and rinsed several times with sterile water. Following stratification for 2 d at 4 °C, the seeds were inoculated onto Type A agar (Sigma, St. Louis, MO, USA) plates with modified 1/10 Hoagland's medium and 1% sucrose and after 1 week, transferred to plates with and without Cd2+. Cd2+ was added to the medium as chloride salt. Plates were incubated vertically under continuous light. After 10–12 d, root length was measured.

Analysis of pH and extractable metals in soil samples

All soil samples were air-dried and sieved through a 1 mm mesh for further analysis. The pH of the soil samples was measured in the mixture of soil and 1 m KCl (1.0:2.5) after 24 h extraction. About 3.0 g of dried soil was extracted with 25 mL of 0.1 m HCl for 30 min at 25 °C. Metal content was analysed by atomic absorption spectroscopy (AAS).

Microarray experiments and data analysis

For each experiment, roots of three A. thaliana and three A. halleri plants each were pooled and homogenized in liquid nitrogen prior to RNA isolation. Total RNA was extracted from roots with TRIzol (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's protocol. Ten micrograms of each RNA sample were used to synthesize first- and second-strand cDNA according to the Affymetrix manual. Biotinylated complementary RNAs (cRNAs) were in vitro transcribed using the MEGAscript T7 Kit (Ambion, Austin, TX, USA). For hybridization of labelled cRNAs, ‘Arabidopsis genome array’ (Affymetrix, Santa Clara, CA, USA), representing more than 8000 A. thaliana genes, were used. Hybridization, staining and scanning of the microarrays were carried out according to the Affymetrix manual. Microarray Suite 5.0 (Affymetrix) was used for data normalization using default settings. The overall intensity of all probe sets of each array was scaled to 500. Further analyses were carried out using GeneSpring 7.0 (Silicon Genetics, Redwood City, CA, USA). Prior to fold-change calculation, data sets were normalized using default settings. All independent experiments were analysed together and defined as replicates. After the normalization, probe sets were filtered out, which were then called ‘present’ or ‘marginal’ in at least 2 independent treatment-experiments. Afterwards, fold-changes were calculated using the fold-change interpretation. Probe sets, which showed at least 2-fold higher (lower) expression signal were called induced (repressed). To identify significant changes, a one-way analysis of variance (anova) was carried out with these probe sets. For the statistical analysis, the default settings of GeneSpring 7.0 were used. Additionally, a statistically less stringent analysis was carried out for the low dose Cd2+-treatment data using Microsoft Excel. For the identification of induced genes, each independent experiment was first analysed separately. Probe sets were filtered out, which were called ‘present’ or ‘marginal’ in the treated sample. Afterwards, the fold-change of each probe set was calculated by dividing the signal intensity of the treated sample by that of the control sample. Probe sets were called induced, when the calculated fold-change was two or higher in two out of three independent experiments (biological replicates).

Real-time quantitative PCR

About 1.5 µg of total RNA were treated with DNAse I (Invitrogen) and used for cDNA synthesis according to the manufacturer's instructions (SuperScript First-Strand Synthesis System for RT-PCR, Invitrogen). Afterwards, the cDNA was diluted with water (1:50). PCR reactions (5.0 µL diluted cDNA, 2.5 pmol of each primer, 10.0 µL SYBR green RT-PCR reagent in a final volume of 20.0 µL) were performed in an ABI Prism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA, USA) using default settings. Each sample was analysed in triplicate and the resulting data were analysed using ABI Prism 7000 SDS Software (Applied Biosystems). For the calculation of the threshold cycle (CT) values, the auto-CT function was used after excluding imperfect samples. For further calculations, the mean value of each triplicate was used. To normalize the target gene expression, the difference between the CT of the target gene and the CT of EF1α (= constitutive control) for the respective template was calculated (= ΔCT value). To be able to calculate the fold-changes in gene expression, the ΔΔCT-value was calculated as follows: ΔΔ CT = ΔCT (treated sample) − ΔCT (control sample). For the calculation of fold-changes we used the formula: x-fold-change = 2ΔΔCT. Primer sequences for the A. halleri genes were derived from cloned cDNA fragments.

Elemental analysis by AAS

After the 2-h treatment, the roots were harvested and washed with Millipore H2O, three times with 10 mm CaCl2 and again with Millipore H2O. All washing steps were carried out at 4 °C for 10 min and under gentle shaking. Samples were dried and dry weight (DW) was determined. The samples were digested in a microwave in 10 mL HNO3/H2O (6 mL 65% HNO3, 4 mL Millipore H2O). Afterwards the volume was adjusted to 25 mL with Millipore H2O. The Cd and Cu contents were analysed by flame and graphite furnace AAS using a Perkin Elmer AAnalyst 800 (Perkin Elmer, Uberlingen, Germany).


Metal tolerance of A. halleri

Metal tolerance of A. halleri plants in comparison to A. thaliana was tested in two ways. Firstly, A. halleri and A. thaliana seedlings were grown in soil taken from sites in Langelsheim (N 51°56.569′, E 10°20.955′) and in the Harz mountains (near Schierke) (N 51°45.413′, E 10°41.147′) where A. halleri is growing, but where the soil is not metal-contaminated (please note that A. halleri is a facultative metallophyte, i.e. it can grow both in metal-rich and in normal soil). Results of the soil analysis are shown in Table 1. The soil from Langelsheim contained Zn, Cu and Cd concentrations that are about 50–120, 7–10 and 30–60 fold higher, respectively, than the average values for normal surface soils (Kabata-Pendias & Pendias 2001), while the soil from Schierke contained Zn, Cd and Cu within or below the average range. As shown in Fig. 1a., A. halleri grew well in both soil types while A. thaliana plants grew more slowly, became chlorotic and showed severe stress symptoms in the metal-rich soil. Secondly, A. halleri seeds collected from the metal-contaminated site and A. thaliana Col-0 seeds were germinated and grown on vertical plates with or without Cd2+ added. As shown in Fig. 1b., A. halleri seedlings are not affected by a Cd2+ concentration of 20 µm that causes severe root-growth inhibition and chlorosis of cotyledons seen in A. thaliana seedlings.

Table 1.  pH and the content of extractable zinc (Zn), cadmium (Cd) and copper (Cu) in analysed soil samples from sites where Arabidopsis halleri is naturally occurring (µg/g−1 DW)
SiteExtractable metal in soilpH
  1. Shown are mean values ± SD, n = 3.

  2. For GPS co-ordinates of the sites see Results section.

  3. DW, dry weight; GPS, global positioning system.

Control site 11.6 ± 2.61 0.3 ± 0.04  2.1 ± 0.945.0
Contaminated site4904.0 ± 93.2422.1 ± 3.83115.7 ± 4.946.4
Figure 1.

Arabidopsis halleri (A.h.) is more metal-tolerant than Arabidopsis thaliana (A.t.). (a) A. halleri cuttings and A. thaliana seedlings were grown in normal lab soil (GS 90 and Vermiculite; 3:1) for 2–3 and 4–5 weeks, respectively, then transferred to non-contaminated and metal-contaminated soil sampled at sites from the Harz mountains (for metal content of the two soil types see Table 1) and cultivated for an additional 2 weeks. During the whole period, the plants were cultivated under short day conditions (8 h light/16 h dark) at 22 °C. (b) Seeds of A. halleri plants growing at a metal-contaminated site and seeds of A. thaliana were germinated on normal medium, after 7 d transferred to plates with (right) and without (left) 20 µm Cd2+, and grown for additional 10 d.

Short-term root Cd accumulation

For microarray experiments, the plants were grown hydroponically so that it was possible to focus the analysis on root tissue, which is directly exposed to metal ions and therefore has to express metal-tolerance mechanisms. Our strategy was to treat plants with Cd2+ concentrations below and above a dose that is affecting growth and to study rapid transcriptional responses. This was done because our main interest is in understanding primary responses to metal ion exposure as opposed to responses to unspecific cellular damage. In addition, we intended to treat A. thaliana and A. halleri with same-effect concentrations of Cd2+ in the sense that comparable intracellular concentrations can be reached. In order to identify these, we measured the uptake of Cd2+ into the roots of hydroponically grown plants after a 2 h exposure to different Cd2+ concentrations. We found that about 2.5-fold higher external Cd2+ concentrations had to be applied to A. halleri roots in order to achieve equal amounts of Cd2+ inside the roots (data not shown). Consequently, A. thaliana roots were treated with 10 µm Cd2+ and A. halleri roots with 25 µm Cd2+. These concentrations were at the threshold at which growth inhibition was observed in modified 1/10 Hoagland's medium. Cd content was 570 ± 70 and 546 ± 36 ng Cd2+ mg−1 DW for A. thaliana and A. halleri roots (n = 3), respectively, at the end of the treatment period. In addition, the plants were treated with 5-fold higher Cd2+ concentrations. In order to compare the effects of Cd2+ treatment to those elicited by treatment with a redox-active metal, we treated plants for 2 h with 10 µm Cu2+. In addition, this allowed to compare responses to excess of an essential and a non-essential metal. We used the concentration of 10 µm Cu2+ for both species because, firstly, there are no indications of Cu2+ hypertolerance in A. halleri (data not shown), and secondly, Cu2+ ions are more toxic than Cd2+ ions in hydroponic culture.

Cross-species hybridization

Two hours after the addition of either Cd2+ or Cu2+ ions to the nutrient medium, roots were harvested, RNA was extracted and – following cDNA synthesis and synthesis of biotinylated cRNA – Affymetrix GeneChips representing about 8300 A. thaliana genes were hybridized. Three plants were pooled for one sample, and three independent experiments were performed (over a span of 14 months). All arrays were normalized to a target intensity of 500. On average, the mean signal intensities were lower for A. halleri arrays. Thus, the scaling factors were on average higher for A. halleri arrays by a factor of 3.6. For chips hybridized with A. thaliana cRNA, an average of 4779 genes (± 389) were called ‘present’ or ‘marginal’ according to Affymetrix nomenclature with P-values of P < 0.04 or P < 0.06, respectively. This number was – similar to previously reported data (Becher et al. 2004; Weber et al. 2004) – significantly lower for chips hybridized with A. halleri cRNA: 2724 (± 309). However, there was no difference in the reproducibility of data between genes present in A. thaliana or A. halleri. When all replicates of the different experiments were plotted against each other, the correlation coefficients for A. thaliana and A. halleri expression data were in the same range of r = 0.97 to 0.99 (data not shown). We restricted the analysis to genes present in at least two out of three experiments (treated samples).

Firstly we identified genes that are responsive to Cd2+ in the two Arabidopsis species. We defined genes as responsive when transcripts were detected as present in two out of three experiments and the signals showed a significant (P < 0.05) twofold or higher difference. For A. thaliana, 111 genes showed higher expression after 2 h of exposure to the high Cd2+ dose, for A. halleri 5 genes. The most strongly induced A. thaliana genes are listed in Table 2 (complete table: Table S1, Supplementary material), and those for A. halleri genes in Table 3. For the lower Cd2+ dose, no transcript showed a > twofold increase in abundance that passed stringent statistical analysis. However, a number of genes (66 in A. thaliana and 34 in A. halleri) were > twofold higher in at least two out of three independent experiments. Cd2+-responsive genes fulfilling this softer criterion are highlighted in grey inTables 2 and 3. About 19 out of 111 Cd2+-responsive genes in A. thaliana and 3 out of 5 corresponding genes in A. halleri are assumed to respond to external Cd2+ particularly sensitively and in a dose-dependent manner. As can be seen inTables 2 and 3, most of these genes are among those with the highest induction factors in response to the higher Cd2+ dose. Some 56 genes were significantly down-regulated after 2 h under Cd2+ stress in A. thaliana roots (Table S2, Supplementary material). For A. halleri, no transcript showing a significant (P < 0.05) down-regulation was found.

Table 2.  Genes significantly (P < 0.05) induced >5-fold after 2 h of exposure to 50 µm Cd2+ in Arabidopsis thaliana roots
AGI codeDescriptionAffy-IDFold-changeP-value
  • Genes that showed an up-regulation also in response to the lower Cd2+ dose of 10 µm (> 2-fold in at least two out of three experiments) are highlighted in grey, genes belonging to the ‘Arabidopsis Cd2+ core response’ (i.e. whose Arabidopsis halleri orthologs were also Cd2+ responsive) are marked by boxes.

  • a

    One of at least two probe sets representing the same gene.

  • b

    A probe set representing two different genes.

  • AGI, Arabidopsis Genome Initiative; Affy-ID, Affymetrix identification.

At1g5405017.4 kDa class III heat-shock protein (HSP17.4-CIII)15404_at79.70.0401
At2g44840Ethylene-responsive element-binding protein, putative20489_at36.20.0483
At2g29460Glutathione S-transferase, putative19640_at31.70.0327
At3g2520023.6 kDa mitochondrial small heat-shock protein (HSP23.6-M)13282_s_at30.70.0327
At3g21680Proton-dependent oligopeptide transport (POT) family protein19762_at29.00.0406
At1g5354017.6 kDa class I small heat-shock protein (HSP17.6C-CI)13276_at26.90.0442
At1g5986017.6 kDa class I heat-shock protein (HSP17.6A-CI)18945_at18.00.0327
At2g26150Heat-shock transcription factor family protein12431_at15.20.0381
At3g28210Zinc finger (AN1-like) family protein16638_at13.20.0327
At5g08780bAt5g08780: histone H1/H5 family protein; At5g08790: no apical meristem   
At5g08790b(NAM) family protein18591_at12.90.0483
At1g74310bAt1g74310: heat-shock protein 101 (HSP101); At1g74300:   
At1g74300bEsterase/lipase/thioesterase/ family protein13274_at12.40.0367
At3g14680ATP-sulfurylase 3 (APS3)15647_s_at12.30.0401
At3g36500Expressed protein17381_at12.30.0406
At5g57560Xyloglucan:xyloglucosyl transferase (TCH4)16620_s_at12.10.0327
At5g1203017.7 kDa class II heat-shock protein 17.6 A (HSP17.7-CII)13277_i_at11.60.0483
At1g27730aZinc finger (C2H2 type) family protein (ZAT10)18217_g_at11.10.0327
At2g2950017.6 kDa class I small heat-shock protein (HSP17.6B-CI)20323_at10.90.0448
At3g23240Ethylene response factor 1 (ERF1)17514_s_at10.60.0483
At1g27730aZinc finger (C2H2 type) family protein (ZAT10)18216_at10.30.0442
At2g37430aZinc finger (C2H2 type) family protein (ZAT11)20619_at10.00.0367
At3g11280a1-aminocyclopropane-1-carboxylate synthase 6/ACC synthase 612892_g_at 9.60.0327
At5g47220Ethylene-responsive element-binding factor 2 (ERF2)16609_at 9.40.0401
At5g59820Zinc finger (C2H2 type) family protein (ZAT12)13015_s_at 9.30.0381
At2g02010Glutamate decarboxylase, putative18508_at 8.40.0406
At5g47230Ethylene-responsive element-binding factor 5 (ERF5)16536_at 8.30.0327
At3g11280a1-aminocyclopropane-1-carboxylate synthase 6/ACC synthase 612891_at 8.10.0327
At2g38470WRKY family transcription factor17303_s_at 7.50.0381
At3g17490Ethylene-responsive element-binding protein, putative16539_s_at 7.30.0464
At3g11280a1-aminocyclopropane-1-carboxylate synthase 6/ACC synthase 616817_s_at 7.00.0327
At3g22710bAt3g22710: cytochrome P450 family protein; At3g22690: cytochrome12790_s_at 6.50.0327
At3g22690bP450 family protein   
At5g04340C2H2 type zinc finger protein15665_at 6.20.0401
At3g11360Zinc finger (C3HC4-type RING finger) family protein (RHA1b)16553_f_at 6.00.0327
At3g27280Calcium-binding EF hand family protein15431_at 6.00.0327
At3g16050Stress-responsive protein, putative16617_s_at 5.90.0401
At1g62300WRKY family transcription factor13115_at 5.60.0327
At1g72930Toll-Interleukin-Resistance (TIR) domain-containing protein18003_at 5.40.0483
At2g37430aZinc finger (C2H2 type) family protein (ZAT11)20620_g_at 5.20.0483
Table 3.  Genes significantly (P < 0.05) induced >two fold after 2 h of exposure to 125 µm Cd2+ in Arabidopsis halleri roots
  • Shown are the AGI codes and annotations of the Arabidopsis thaliana orthologs. Genes that showed a > twofold up-regulation also in response to the lower Cd2+ dose of 25 µm in at least two out of three experiments are highlighted in grey, genes belonging to the ‘Arabidopsis Cd2+ core response’ (i.e. whose A. thaliana orthologs were also Cd2+ responsive) are marked by boxes.

  • a

    A probe set representing two different genes.

  • AGI, Arabidopsis Genome Initiative; Affy-ID, Affymetrix identification.

At5g47230Ethylene-responsive element-binding factor 5 (ERF5)16536_at5.20.0402
At4g27280Calcium-binding EF hand family protein15431_at3.90.0402
At5g04340C2H2 type zinc finger protein15665_at3.00.0402
At5g67300aAt5g67300 myb family transcription factor; At5g67310 cytochrome P450   
At5g67310afamily protein19707_s_at2.40.0402
At4g14680ATP-sulfurylase 3 (APS3)15647_s_at2.10.0437

In order to be able to compare the Cd2+ responses to those of a redox-active metal we also analysed roots exposed to 10 µm Cu2+ for 2 h. Results were filtered in the same way as was done for the Cd2+ samples. Overall, the number of Cu2+-responsive genes was much higher for both species. About 708 and 165 genes were up-regulated in A. thaliana and A. halleri roots, respectively. In addition, the induction factors were far more extreme than determined for Cd2+-treated roots even though the Cu2+ concentration in the growth medium was only raised to 10 µm for 2 h. Table 4 shows the genes with an induction > 50-fold in A. thaliana roots plus the data for the respective A. halleri orthologs. The complete lists are shown in Tables S3 and S4 (Supplementary material). For 100 genes (probe sets), a more than 20-fold higher transcript level was found in A. thaliana, and for 9 genes in A. halleri. The most extreme example was a protein kinase (At1g70130) in A. thaliana (with an average fold-change of 1179) and the homolog of a putative ethylene response element-binding protein (EREBP)(At2g44840) in A. halleri (with an average fold-change of 90). The latter was also the gene with the third-highest induction factor (490) in A. thaliana. For a total of 359 A. thaliana genes, a significant reduction in abundance in Cu2+-treated plants was determined (Table S5, Supplementary material). Again, no A. halleri transcript showing a significant (P < 0.05) down-regulation was found.

Table 4.  Shown are the Arabidopsis thaliana genes which showed the strongest up-regulation in roots upon 2 h exposure to 10 µm Cu2+ (significantly induced > 50-fold, P < 0.05) plus the data for the respective Arabidopsis halleri orthologs
  • Highlighted in grey are genes that showed significant up-regulation also in A. halleri.

  • a

    One of at least two probe sets representing the same gene.

  • AGI, Arabidopsis Genome Initiative; Affy-ID, Affymetrix identification; A.h., Arabidopsis halleri; A.t., Arabidopsis thaliana.

At1g70130Lectin protein kinase, putative12396_at1179.00.0012 4.8
At2g29460Glutathione S-transferase, putative19640_at1109.00.002412.3
At2g44840Ethylene-responsive element-binding protein, putative20489_at 489.80.015290.9
At1g17420Lipoxygenase, putative18668_at 438.10.031718.9
At3g23250myb family transcription factor (MYB15)12737_f_at 432.30.0248 8.2
At5g20230Plastocyanin-like domain-containing protein19178_at 403.00.005643.4
At3g46090aZinc finger (C2H2 type) family protein (ZAT7)15778_at 333.10.0030 0.5
At2g02010Glutamate decarboxylase, putative18508_at 298.60.0039 4.0
At3g46090aZinc finger (C2H2 type) family protein (ZAT7)15779_g_at 279.50.0012 6.5
At1g19670Coronatine-responsive protein/coronatine-induced protein 1 (CORI1)12916_at 186.40.0220 2.3
At4g23250Protein kinase family protein20246_s_at 183.30.0182 4.6
At2g29450Glutathione S-transferase16082_s_at 150.30.0014 6.6
At4g34410AP2 domain-containing transcription factor, putative19108_at 144.40.0213 4.1
At1g24140Matrixin family protein13842_at 121.90.0257 5.6
At2g22760Basic helix-loop-helix (bHLH) family protein19803_s_at 113.40.0209 3.1
At4g24350Phosphorylase family protein15522_i_at 107.80.0317 3.3
At4g17090aBeta-amylase (CT-BMY)/1,4-alpha-D-glucan maltohydrolase18669_at 105.80.0170 0.3
At2g24850Aminotransferase, putative17008_at 98.10.0058 6.8
At4g11370Zinc finger (C3HC4-type RING finger) family protein16130_s_at 96.60.0419 0.4
At3g50930AAA-type ATPase family protein15424_at 93.50.011215.2
At5g57560Xyloglucan:xyloglucosyl transferase (TCH4)16620_s_at 89.50.001246.0
At4g14680ATP-sulfurylase 3 (APS3)15647_s_at 86.50.0020 4.2
At2g32140Disease resistance protein (TIR class), putative13312_at 79.20.0328 0.8
At4g17090aBeta-amylase (CT-BMY)/1,4-alpha-D-glucan maltohydrolase18670_g_at 79.20.0324 0.2
At2g28210Carbonic anhydrase family protein14461_at 77.50.0239 0.8
At2g14290F-box family protein16272_at 77.20.0184 1.5
At2g02340F-box family protein17157_at 76.70.0066 3.4
At2g34930Disease resistance family protein contains leucine rich-repeat domains12251_at 76.40.0164 4.3
At2g18210Expressed protein18258_at 74.40.0052 4.2
At3g28210Zinc finger (AN1-like) family protein16638_at 70.80.0055 6.3
At2g32210Expressed protein18267_at 69.70.012426.9
At1g08860Copine, putative17134_at 67.20.003823.9
At4g11280a1-aminocyclopropane-1-carboxylate synthase 6/ACC synthase 612892_g_at 65.80.0033 5.6
At4g25810Xyloglucan:xyloglucosyl transferase, putative17533_s_at 62.10.001237.7
At2g35980Harpin-induced family protein (YLS9)19991_at 61.20.006245.1
At4g36500Expressed protein17381_at 59.00.001222.4
At5g59820Zinc finger (C2H2 type) family protein (ZAT12)13015_s_at 58.30.0036 4.0
At2g42360Zinc finger (C3HC4-type RING finger) family protein14320_at 57.20.0017 5.4
At4g11280a1-aminocyclopropane-1-carboxylate synthase 6/ACC synthase 612891_at 55.80.0056 6.9
At4g15100Serine carboxypeptidase S10 family protein18132_at 55.60.0487 5.9
At3g23240Ethylene response factor 1 (ERF1)17514_s_at 54.70.0025 3.9
At3g15356Legume lectin family protein18228_at 50.40.001211.5

Verification of microarray results

We selected four genes from A. thaliana and their respective orthologs from A. halleri with contrasting expression patterns for quantitative real-time PCR analysis to verify the microarray data. The small heat-shock protein gene At1g53540 or its respective A. halleri ortholog were up-regulated in response to both low (L-Cd) and high (H-Cd) Cd2+ concentrations in both species, yet was Cu2+-responsive only in A. thaliana. The transcript of At4g28350 and its A. halleri ortholog, encoding receptor-like kinases, appeared to be more abundant upon exposure to low Cd2+ in A. thaliana and upon exposure to Cu2+ in both species. At2g44840 and its A. halleri ortholog, encoding an ethylene-response element-binding protein, were responsive to all treatments in both species yet to very different degrees. Finally, the blue copper-binding protein gene At5g20230 or its A. halleri ortholog were induced by Cd2+ and Cu2+ in A. thaliana but only by Cu2+ in A. halleri. Real-time PCR was performed for these eight genes in two independent experiments. The results, presented in Fig. 2 and Table 5, showed very good agreement with the microarray data both qualitatively and quantitatively. More importantly, the microarray data were confirmed equally well for both species, demonstrating again the reliability of the data obtained by cross-species hybridizations. Responsiveness or non-responsiveness of a gene to a particular treatment was confirmed in 23 out of 24 cases, the only exception being the small heat-shock protein in A. halleri, for which a slight induction in response to the high Cd2+ concentrations was found only by real-time PCR. In addition, the examples of At1g53540, At2g44840 and At5g20230 demonstrate that the softer selection criterion for determining genes responsive to the lower Cd2+ dose was justified. Increase in transcript abundance was not significant according to the stringent criteria, yet the up-regulation detected in two out of three array experiments could be confirmed independently by quantitative real-time PCR.

Figure 2.

Verification of array results by quantitative real-time PCR. Four genes from Arabidopsis thaliana (left) and their orthologs from Arabidopsis halleri (right) with contrasting expression patterns were selected. Primers were derived from the known A. thaliana sequences and from cloned and sequenced fragments of the respective A. halleri orthologs. Plants were grown and treated as those used for the microarray analysis. The summary of two independent experiments is shown. All samples were analysed in triplicate. Low (L-Cd) and high (H-Cd) Cd2+ concentrations, threshold cycle (CT).

Table 5.  Comparison of average fold-changes derived from microarray and quantitative real-time PCR data for four selected Arabidopsis thaliana genes and their respective Arabidopsis halleri orthologs that showed contrasting expression patterns
GeneTreatmentA. thalianaA. halleri
  1. MA fold-changes derived from Genespring analysis were log2 transformed. qRTPCR: ΔΔCT-values are listed.

  2. MA, Microarray data; qRTPCR, quantitative real-time PCR; L-Cd, low Cd2+ concentration; H-Cd, high Cd2+ concentration; CT, threshold cycle.

sHSP (At1g53540)L-Cd2.6 2.1−0.30.7
H-Cd4.7 5.9 0.02.3
Cu4.2 4.9 0.50.2
EREBP (At2g44840)L-Cd0.8 1.0 4.53.4
H-Cd5.2 6.2 4.74.8
Cu8.9 9.1 6.56.9
BCBP (At5g20230)L-Cd2.2 1.6 0.60.6
H-Cd2.2 3.2 0.10.6
Cu8.710.0 5.47.5
RLK (At4g28350)L-Cd0.3 0.2 0.31.1
H-Cd1.5 1.3 1.01.0
Cu5.6 5.9 4.26.3

Comparative analysis of the array data

A major objective of this study was to gain insight into excess metal responses by comparing a normal plant and a metallophyte. Sorting of the genes up-regulated after 2 h in the presence of Cd2+ in A. thaliana into functional categories (according to the The Arabidopsis Information Resource (TAIR) catalogue,, revealed that three categories were strongly overrepresented: abiotic- and biotic-stress response (5.13 and 2.60% of the responsive genes and of the genes represented on the array, respectively), general stress response (4.11 and 2.06%, respectively) and transcription (5.95 and 3.60%, respectively). Underrepresented were transport (3.90 and 5.64%, respectively), electron transport or energy pathways (1.23 and 2.37%, respectively) and DNA and RNA metabolism (0.62 and 1.22%, respectively). Most of the genes with high induction factors encode small heat-shock proteins and overall, the annotation of at least 15 genes directly refers to heat shock. Other typical stress-response proteins in Table 2 and Table S1 include GS transferases (e.g. At2g29460, At4g17500) and peroxidases (At5g39580). In addition, a large number of transcription factor genes and genes related to the synthesis of or response to ethylene are among the list of Cd2+-responsive genes. An expected up-regulation of sulphate assimilation (Heiss et al. 1999) is indicated by the increased transcript abundance of an ATP sulfurylase (At4g14680) and the two 5′-adenylylsulfate reductases APR1 and APR2 (At4g04610 and At1g62180, respectively). One of the genes with the strongest response encodes a proton-dependent oligopeptide transport (POT) family protein (At4g21680). This might be relevant because of the importance of metal-binding peptides (PCs) for Cd2+ tolerance (Rea et al. 2004).

In spite of comparable internal Cd2+ concentrations in Cd2+-treated roots, the number of Cd2+-responsive genes detected in A. halleri in relation to the total number of transcripts present was lower for A. halleri by about a factor of 10. Of the five Cd2+-responsive genes in A. halleri, four are up-regulated also in A. thaliana(Fig. 3). They constitute the ‘Arabidopsis Cd2+ core response’ and encode the ATP sulfurylase 3, an EF-hand containing putative Ca2+-binding protein (At4g27280) and two putative transcription factors (At5g04340 and At5g47230). The sole probe set that indicated an A. halleri-specific response to Cd2+ is ambiguous and belongs to an R2R3-MYB transcription factor homologue (At5g67300) and to a cytochrome P450 family member (At5g67310).

Figure 3.

Venn diagrams displaying the overlaps between Cu2+ and Cd2+ responses in Arabidopsis thaliana and Arabidopsis halleri (a) as well as the overlaps in responses between the two species (b). The diagrams in (a) reveal the existence of Cd2+-specific responses. There are 4 and 140 genes that constitute the ‘Arabidopsis Cd2+’ and the ‘Arabidopsis Cu2+’ core responses, respectively (b).

The magnitude of excess Cu2+ effects on the root transcriptome of the two species in terms of number of responsive genes was less different than found for Cd2+. The ratio of up-regulated genes to detectable genes was only about 2.4-fold higher for A. thaliana (14.8 versus 6.1%) instead of > 10-fold in the case of Cd2+. For both species, the three overrepresented functional categories among up-regulated genes were abiotic- and biotic-stress response (5.38 and 2.6% for A. thaliana[A.t.], 4.49 and 2.6% for A. halleri[A.h.]), general stress response (4.11 and 2.06% for A.t., 3.91 and 2.06% for A.h.) and signal transduction (1.94 and 1.53% for A.t., 2.32 and 1.53% for A.h.). Most of the genes showing the strongest responses in both species encode either known and suspected stress proteins such as GS transferases, a lipoxygenase, a plastocyanin-like domain-containing protein, TCH4 or transcription factors belonging to various classes (e.g. AP2/EREBP, MYB, C2H2, bHLH). In addition, among the genes showing the strongest Cu2+-dependent up-regulation were again APS3 and the POT family member.

For 87% of the genes up-regulated under excess Cu2+ in A. halleri (140 out of 165), the respective A. thaliana ortholog was also Cu2+-responsive (Fig. 3). Thus, the ‘Arabidopsis Cu2+ core response’ is largely identical to the list shown in Table S4 and consequently there is little difference in the representation of functional categories. In addition, there was a good agreement in the relative strength of the Cu2+ responses in the two species. For instance, for 11 out of the top 15 genes with the highest fold increase and 18 out of the 31 genes induced by Cu2+ in A. halleri roots more than 10-fold, the respective A. thaliana orthologs were among the 100 genes showing the strongest Cu2+-elicited up-regulation in A. thaliana.

Previous experiments had shown that a number of genes are constitutively much more strongly expressed in A. halleri roots relative to A. thaliana roots (Weber et al. 2004). Thus, a potentially interesting group of genes are those responsive to Cd2+ or Cu2+ in A. thaliana and constitutively higher in A. halleri. We extended the initial list of strongly expressed A. halleri genes (Weber et al. 2004) by including additional control samples for the constitutive comparison. Furthermore, transcripts present only in A. halleri were taken into account. The resulting list of 48 genes showing a statistically significant P value (P < 0.05) and at least a 7-fold higher transcript level in A. halleri roots relative to A. thaliana roots is shown in Table S6 (Supplementary material). The cut-off of 7 is derived from a multiplication of the average scaling-factor difference between A. thaliana and A. halleri hybridizations (see above) with the common cut-off of 2. No functional category other than metal-homeostasis factors (which is not a gene ontology category) is particularly overrepresented among these genes. Five A. thaliana genes and their A. halleri counterparts fulfil the criterion of being metal-responsive in A. thaliana and constitutively higher expressed in A. halleri. One of these is responsive to both Cu2+ and Cd2+ and encodes a putative receptor-like kinase (At1g78830). Three genes are Cu2+-responsive in A. thaliana. They code for a cytochrome P450 monooxygenase (At2g30750), a protein with similarity to bacterial S-adenosyl-methionine (SAM)-dependent methyltransferases (At2g41380) and a member of the major facilitator superfamily (At2g16660). Finally, the Cd2+-responsive A. thaliana gene is AtZIP9 (At4g33020) (Table 6).

Table 6.  Genes that are responsive to Cu2+ and/or high Cd2+ in Arabidopsis thaliana roots and whose orthologs are constitutively highly expressed in Arabidopsis halleri roots
AGI-codeDescriptionAffy-IDA.t. Cd
A.t. Cu
A.t. versus A.h.
  • Shown are the average fold-changes (n = 3) upon Cd2+ or Cu2+ treatment in A. thaliana roots and the fold difference between root expression levels for the A. thaliana genes and their A. halleri orthologs. Highlighted in grey are significant (P < 0.05) differences (2-fold for the metal treatments, 7-fold for the constitutive comparison).

  • a

    A probe set representing two different genes.

  • AGI, Arabidopsis Genome Initiative; Affy-ID, Affymetrix identification; A.t., Arabidopsis thaliana; A.h., Arabidopsis halleri.

At4g33020Metal transporter, putative (ZIP9)14831_at2.1 0.822.8
At2g41380Embryo-abundant-related protein14083_at0.913.010.4
At2g16660Nodulin family protein20190_at0.816.5 9.3
At2g30750aCytochrome P450, putative    
At2g30770aCytochrome P450, putative14609_at1.3 3.9 8.2
At1g78830aCurculin-like (mannose-binding) lectin family protein    
At1g78820aCurculin-like (mannose-binding) lectin family protein12817_g_at2.918.8 7.0

A second principal comparison is that between the responses to the ions of the redox-active metal copper and the non-redox-active metal cadmium (Fig. 3). Given the strong response to excess Cu2+, it was of particular interest to identify genes specifically responsive to Cd2+ because these genes might represent good candidates for being involved in the primary response of plant roots to Cd2+ exposure. Table 7 lists the genes which showed up-regulation under Cd2+ exposure yet were not changed significantly in Cu2+-treated roots. For A. thaliana, this list comprises 23 genes, for A. halleri it is just one probe set representing 2 different genes, the ortholog of the A. thaliana R2R3-MYB transcription factor (At5g67300) and a cytochrome P450 family member (At5g67310). This was also identified as the only A. halleri-specific Cd2+ response in the cross-species comparison (see above, Table 3). Among the 23 Cd2+-specific A. thaliana genes, there are 6 putative transcription factor genes, at least 2 genes encoding putative kinases (At1g53700, At2g23770), and again the metal-transporter gene AtZIP9. The ‘Arabidopsis Cd2+ core response’ defined above is identical to the ‘Arabidopsis excess metal core response’. All four genes responsive to the Cd2+ treatment in both species were also responsive to excess Cu2+ in both species.

Table 7.  Genes specifically induced by Cd2+ and not by Cu2+ in Arabidopsis thaliana roots
  1. n = 3, P < 0.05.

  2. AGI, Arabidopsis Genome Initiative; Affy-ID, Affymetrix identification.

At3g16050Stress-responsive protein, putative16617_s_at5.9
At1g53700Protein kinase, putative16790_at4.8
At5g24090Acidic endochitinase (CHIB1)19455_s_at3.1
At2g32560F-box family protein15529_at3.0
At4g31000Calmodulin-binding protein19857_at2.9
At1g04310Ethylene receptor-related protein12444_s_at2.8
At1g67360Rubber elongation factor (REF) family protein19944_at2.7
At5g16600myb family transcription factor (MYB43)12706_f_at2.5
At2g15830Expressed protein14090_i_at2.5
At1g14350myb family transcription factor (MYB124)17971_s_at2.5
At4g11660Heat-shock factor protein 7 (HSF7)19629_at2.5
At2g41800Expressed protein14517_at2.4
At2g28710Zinc finger (C2H2 type) family protein16198_at2.3
At5g16600myb family transcription factor (MYB43)17608_at2.3
At4g07960Glycosyl transferase family 2 protein17614_at2.3
At2g46780RNA recognition motif (RRM)-containing protein12546_at2.2
At3g46130myb family transcription factor (MYB48)12709_f_at2.2
At3g16770AP2 domain-containing protein RAP2.313240_f_at2.2
At2g45920U-box domain-containing protein19175_at2.2
At2g23770Protein kinase family protein20123_at2.2
At4g33020Metal transporter, putative (ZIP9)14831_at2.1
At4g01630Expansin, putative (EXP17)19346_at2.1
At3g51920Calmodulin-9 (CAM9)17990_at2.0


With this study we aimed at exploiting comparative transcriptome analysis to elucidate the basis of metal hypertolerance in metallophytes, to learn more about modes of toxicity of Cd2+ ions in comparison to Cu2+ and to identify candidate genes potentially involved in mediating transcriptional responses to excess of metal ions. We assumed that a comparative analysis might reveal otherwise not obtainable valuable information both by extracting the responses shared across the species as well as by identifying differential responses.

The main conclusions and hypotheses, extractable from the data even though they cover only about one-third of the A. thaliana genome, are: (1) we found little evidence for oxidative stress elicited by Cd2+, instead, short-term Cd2+ exposure causes an apparent Zn2+ deficiency and an increase in the number of misfolded proteins; (2) the Cd2+ hypertolerance of A. halleri is, at least in part, due to a lower short-term Cd2+ uptake rate and a more efficient sequestration of Cd2+ ions in root cells; (3) contributing to the adaptation of A. halleri to metal-rich soil might be the constitutive high expression of genes that are Cd2+-responsive in A. thaliana; (4) there are distinct signalling cascades specifically activated by Cd2+ ions; and (5) comparative transcriptome analysis by cross-species hybridization of oligonucleotide arrays is robust and produces reliable results.

Reduced Cd2+ uptake rates and a more efficient sequestration in root cells contribute to Cd2+ hypertolerance of A. halleri

Based on the field observation that A. halleri hyperaccumulates zinc and cadmium (Dahmani-Muller et al. 2000; Bert et al. 2002), one can postulate a hypertolerance towards the ions of these metals and for some A. halleri populations this has been shown relative to Arabidopsis lyrata (Bert et al. 2003). Growth assays such as those shown in Fig. 1 demonstrated that the A. halleri plants we obtained from a metal-rich site are indeed substantially more metal-tolerant and specifically more Cd2+-tolerant than A. thaliana.

Two factors contribute to this Cd2+ hypertolerance. In hydroponic cultures, A. halleri roots took up significantly less Cd2+ than A. thaliana roots in short-term experiments (2 h). Furthermore, when concentrations were applied that resulted in a comparable root-Cd content, the effect on the A. halleri root transcriptome was still much smaller than on the A. thaliana transcriptome. This indicates that Cd2+ ions are more efficiently separated from Cd2+-sensitive sites in A. halleri root cells. The underlying mechanism is molecularly unknown. Possible candidate genes for the more efficient sequestration of Cd2+ are those which are responsive to metal excess in A. thaliana and constitutively more strongly expressed in A. halleri. However, there is no functional data available yet on the putative Zn2+/Cd2+ transporter ZIP9, the putative facilitator At2g16660 or any other gene listed in Table 6. It will be interesting to see whether mimicking the A. halleri expression of these genes will result in a higher Cd2+ and/or Cu2+ tolerance in A. thaliana.

The constitutive high expression of certain stress-response genes in plants able to thrive in a particular stress environment emerges as a possibly widespread adaptive mechanism. This would imply that it is not so much the expression of particular species-specific stress tolerance genes, but rather the altered regulation of conserved genes that enable certain plants to survive in harsh environments. For the salt cress Thelungiella halophila, for instance, a salt-tolerant relative of the glycophyte A. thaliana, it was shown by comparative transcriptome analysis that the orthologs of several well-known A. thaliana stress-response genes are highly expressed even under control conditions (Taji et al. 2004). Similarly, A. halleri was found previously to express a number of metal-homeostasis genes at much higher levels than A. thaliana independent of micronutrient status (Becher et al. 2004; Weber et al. 2004).

Protein denaturation and apparent Zn deficiency are immediate toxic effects of Cd2+ exposure

Despite the recently growing interest in toxicogenomics (Waters & Fostel 2004), there is currently very little information on global responses to metal excess and on what the primary effects of metal excess are. Our data on the rapid response of A. thaliana roots to Cd2+ suggest modes of toxicity of this non-essential metal. We obtained little evidence for oxidative stress as a primary effect of Cd2+ exposure. Instead, we found a dominance of genes encoding small heat-shock proteins and heat-shock transcription factors. Among the 20 genes showing an increase in transcript abundance of 10-fold or higher, 7 encode heat-shock proteins and 1 encodes a heat-shock transcription factor. About 6 of these also belong to the group of 16 genes apparently responsive already to the low Cd2+ dose in A. thaliana. It has been known for some time that small heat-shock proteins are induced under metal excess as well as under a variety of other stress conditions (Sun, Van Montagu & Verbruggen 2002), and this induction indicates an increase in the number of denatured proteins. Such an effect is to be expected given the strong affinity of Cu2+ and Cd2+ ions to various functional groups in proteins.

It has been hypothesized that one toxic effect of Cd2+ ions is attributable to their competition with Zn2+ ions (Stohs & Bagchi 1995). Our array data indicate that Cd2+ exposure might indeed very rapidly result in apparent Zn deficiency. AtZIP9, hypothesized to encode a Zn2+ uptake system and known as a marker of Zn deficiency (Wintz et al. 2003), is among the genes that are specifically induced by Cd2+ in A. thaliana. Because it is unlikely that a 2 h exposure would suffice to cause actual Zn deficiency, that is, a significant decrease in bulk Zn levels, the up-regulation of AtZIP9 rather suggests that a Zn2+-sensing molecule is occupied by Cd2+ ions as a consequence of the Cd2+ exposure. A similar result was described for S. pombe cells (Chen et al. 2003). Exposure of cells to Cd2+ led to an up-regulation of a ZIP gene (SPBC16D10.06) within 15 min. In addition, similar to A. thaliana, other environmental stress treatments (oxidative stress, heat shock, osmotic stress, DNA damage) did not affect expression level of this gene.

Cd2+-specific responses of A. thaliana roots provide evidence for distinct signalling cascades activated by Cd2+ exposure

Concerning frequently reported transcriptional responses of an organism upon exposure to toxic non-essential metal ions (or to any other non-physiological toxin), at least two fundamental questions need to be answered. Firstly, is there a specific response to the metal ion even though it apparently has no biological function, or is the response rather to damage occurring as a consequence of the exposure? The former would necessarily imply a specific sensing mechanism. Secondly, which genes and proteins are involved in the signalling networks mediating the activation or down-regulation of genes? Concerning the specificity of responses to a particular stimulus, the availability of large data sets derived from close to 2000 microarray experiments represent an invaluable tool (Zimmermann et al. 2004). We analysed the different gene classes extracted from our data against the Genevestigator database. Two hours of Cu2+ treatment resulted in a massive up-regulation of a large number of genes. The overlap between A. thaliana and A. halleri was extensive. Most of the genes responsive to Cu2+ in A. halleri were also found in the A. thaliana list. With very few exceptions, the genes of this ‘Arabidopsis Cu2+ core response’ are strongly responsive to many other abiotic stresses such as ozone, salt, cold and osmotic shock. Thus, we did not find any evidence for Cu2+-specific responses. This is likely due to the fact that excess Cu2+ triggers the massive generation of reactive oxygen species – which is a consequence of most other biotic and abiotic stresses (Mittler 2002; Jonak, Nakagami & Hirt 2004).

Because of the good representation of general stress-response genes on the list of Cu2+-induced genes, a comparison to this list is a suitable way to search for Cd2+-specific genes. Remarkably, 23 transcripts were identified as being specifically up-regulated in A. thaliana under Cd2+ treatment in our experiments. According to data in the Genevestigator database, seven of these are not up-regulated (i.e. showing an expression ratio > twofold) by any other biotic- or abiotic-stress condition tested so far. The other genes are mostly induced by senescence, Agrobacterium infection or high salt. None of them shows an induction pattern reminiscent of those found for the Cu2+-responsive general stress genes. The majority of these specifically Cd2+-responsive genes in A. thaliana roots encode putative signal-transduction components. Future reverse-genetics studies will show if any of them are essential for full metal tolerance and/or required for mediating transcriptional responses to Cd2+.

The detection of apparently highly Cd2+-specific transcriptional responses suggests either the existence of a specific Cd2+-sensing system or the perception of rather specific Cd2+ effects. Such an effect could be the interference with Zn2+ sensing as discussed earlier. It will therefore be interesting to see whether any of the genes discussed here are up-regulated in Zn-deficient A. thaliana roots. A second rather specific Cd2+ effect could be a strain on the GSH pool due to formation of bis(glutathionato)Cd complexes (Li et al. 1997) and PC synthesis. Sulphate assimilation is indeed activated (APS3 is part of the Cd2+ core response), yet apparently in a manner different from responses to sulphur starvation. Under sulphur-depleting conditions, there is little evidence for APS up-regulation (Kopriva & Koprivova 2004). The difference in responses implies the existence of distinct signalling cascades, as suggested by Heiss et al. (1999), depending on whether there is sulphur starvation or a strong intracellular sink for sulphur.

In conclusion, we hypothesize, based on our array data, that in plants, specific signalling in response to Cd2+ exposure exists also. The expression of highly specific Cd2+ sensors is well documented in bacteria. ArsR-SmtB repressor family proteins such as CadC from Staphylococcus aureus and CmtR from Mycobacterium tuberculosis repress the expression of Cd2+-efflux pumps (Busenlehner, Pennella & Giedroc 2003; Cavet et al. 2003). The DNA-binding activity is strongly reduced upon metal binding. In S. cerevisiae, the observation that the sulphur–amino acid biosynthesis pathway is activated by Cd2+ and not by other abiotic-stress treatments led Vido et al. (2001) to speculate about a very specific control mechanism for Cd2+ responses. Indeed, recent studies both with S. cerevisiae and S. pombe provide evidence for and mechanistic insight into Cd2+-specific metabolic regulation mediated by ubiquitin–ligase pathways (Barbey et al. 2005; Harrison et al. 2005). Cd2+-sensing proteins, however, are not known yet from yeast. Metal-sensing transcriptional regulators described to date specifically bind Zn (Zap1), Fe (Aft1) or Cu ions (MAC1, ACE1) (Winge, Jensen & Srinivasan 1998). In plants, no metal-sensor protein has been identified yet. Likewise, downstream events translating a change in metal status or specifically the presence of supraoptimal or toxic metal-ion concentrations into modulation of gene expression, mRNA or protein stability are largely unknown. There is evidence for a phosphorylation dependence of transcriptional Cd2+ responses (Suzuki, Koizumi & Sano 2001). Well-known stress-responsive mitogen-activated protein (MAP) kinases might be involved in phosphorylation cascades as they are activated also under Cu2+ and Cd2+ excess (Jonak et al. 2004). A distinction between specific and damage-induced responses, however, is not possible based on these studies.

Plant hormones are often involved in responses to both biotic and abiotic stimuli. We therefore tested growth of well-known signalling mutants in the presence of Cd2+. However, neither SA-deficient nahG plants nor the mutants etr1-1 and ein2-1 appeared to be compromised significantly in their Cd2+ tolerance (data not shown).


We thank Silke Bergmann, Ines Volkmer, Christoph Hutter and Dr Martin Staege for their help with the microarray experiments. We are grateful for the expert technical assistance of Marina Häussler. This work was financially supported by the Deutsche Forschungsgemeinschaft (CL 152/3-1) and in part by the European Union (METALHOME project HPRN-CT-2002-00243).

Supplementary Material

The following supplementary material is available for this article online:

Table S1. Genes significantly (p < 0.05) induced > 2 fold after 2 hrs of exposure to 50 µM Cd2+ in A. thaliana roots.

Table S2. Genes significantly (p < 0.05) down-regulated (< 0.5 fold) after 2 hrs of exposure to 50 µM Cd2+ in A. thaliana roots.

Table S3. Genes significantly (p < 0.05) induced > 2 fold after 2 hrs of exposure to 10 µM Cu2+ in A. thaliana roots.

Table S4. Genes significantly (p < 0.05) induced > 2 fold after 2 hrs of exposure to 10 µM Cu2+ in A. thalleri roots.

Table S5. Genes significantly (p < 0.05) down-regulated (< 0.5 fold) after 2 hrs of exposure to 10 µM Cu2+ in A. thaliana roots.

Table S6. Genes showing a statistically significant (p < 0.05) at least 7 fold higher transcript level in A. halleri roots relative to A. thaliana roots.

Supplementary material is available to download as part of the full-text article from