Mercuric-ion-induced gene expression was studied in Arabidopsis thaliana Columbia wild type. Rosettes of plants grown for 21 d on agar medium supplemented with 20, 30 and 40 µm HgCl2 were pooled and used to isolate cDNAs of induced genes by suppression subtractive hybridization. Of the 576 clones isolated initially, 31 turned out to be mercury-induced by Northern hybridization. However, kinetic studies using cDNA arrays clearly showed that seven genes were exclusively mercuric-ion-induced, 14 were induced by mercury but also affected by a diurnal rhythm, and 10 clones were only modulated by the day–night cycle. The expression levels of the metal-induced genes increased from 1·5-fold to 10-fold. Functional classification resulted in genes encoding proteins for the photosynthetic apparatus and for the antioxidative system. In addition, unexpected genes, whose connection to mercury ion stress is not evident, were identified.
Mercury pollution is one of the world’s most serious environmental problems (Pilon-Smits & Pilon 2000). Because of its chemical properties – it exists as an elemental metal, mercuric ions and organomercurials – mercury can be found in nearly every environmental compartment. As well as the emission of elemental mercury into the atmosphere, mercuric ions are distributed throughout aquatic and soil environments (Leonard et al. 1998). The biochemical toxicity of mercuric ions is conditioned by their reaction with sulfhydryl groups of proteins and by the induction of the Fenton reaction, resulting in the production of reactive oxygen species (ROS) (Ernst 1996). In plants, mercuric ions are detoxified by phytochelatins or by their precursor glutathione, both of which can bind mercuric ions to sulfhydryl groups (Zenk 1996). The problem of mercury contamination in relation to other heavy metals or organic compounds can be solved by phytoremediation (Salt et al. 1995; Meagher 2000). Transgenic plants that harboured the bacterial mer operon (thus expressing an organomercurial lyase and mercuric reductase) were created for the biotechnological soil detoxification of organomercurials and mercuric ions (Rugh et al. 1996; Salt, Smith & Raskin 1998; Bizily, Rugh & Meagher 2000). This resulted in the production of volatile mercury, which is not thought to pose a threat to the environment (Meagher 2000). Apart from this biotechnological approach, there are few studies of the effects of mercuric ions on plants. Short-term plant or cell culture exposure to Hg2+ resulted in a transcriptional increase in both pathogenesis-related proteins (Malehorn, Scott & Shah 1993; Margis-Pinheiro et al. 1993; Wu, Kriz & Widholm 1994; Didierjean et al. 1996) and heat shock proteins (Wollgiehn & Neumann 1999). Such a transcript accumulation has also been demonstrated for other environmental stress factors and it is assumed that ROS are involved in the signalling process (Langebartels et al. 2000 and articles therein). Similarly, mercury-induced oxidative stress, resulting in increased antioxidant enzyme activity, has been described in tomato seedlings (Cho & Park 2000).
On contaminated soil sites, the stress is permanent (in contrast to episodic environmental stresses such as ozone or increased UV-B radiation), regardless of the heavy metal pollutant. Therefore, long-term molecular responses rather than short-term exposure responses have to be taken into account. Transient increases in several mercuric-ion-induced transcripts were found in maize leaves sprayed with a mercuric chloride solution (Didierjean et al. 1996). Similarly, stress-induced cDNA libraries have been constructed from short-term exposed plant material. In Arabidopsis thaliana, long-term growth on mercuric chloride resulted in an enhanced level of metallothionein mRNA, whereas pathogenesis-related protein transcripts could not be detected (Heidenreich et al. 1999). In this study, we used the model plant Arabidopsis thaliana to investigate transcriptional gene activation during long-term mercuric ion exposure. Suppression subtractive hybridization (SSH) (Diatchenko et al. 1996) was performed to clone cDNA fragments of differentially expressed genes.
This study sheds light upon the biochemical response of plants to mercury exposure and the specifically induced cDNAs may be of future use as markers for mercury stress.
Materials and methods
Arabidopsis thaliana ecotype Columbia wild type seeds were surface-sterilized as described by Ohl et al. (1990) and sown on MS agar (pH 5·9) containing sucrose (3% w/v) (Sigma, Deisenhofen, Germany). The agar contained 0, 20, 30 or 40 µm mercuric chloride. After cold treatment in the dark for 3 d at 8 °C, the seeds were transferred into a controlled environment cabinet and seedlings were grown for 21 d under a 14/10 h day–night cycle with a photosynthetic photon flux density of 116 µmol m−2 s−1 from 0700 to 2100 h middle European summer time (MET) and a temperature of 22/20 °C. For the determination of diurnal transcript modulation, plants were grown on standard soil (type P; Fruhstorfer, Lauterbach, Germany) and cultivated as described earlier. Leaf rosettes were frozen in liquid nitrogen, homogenized with mortar and pistil and stored at −80 °C for further analysis.
RNA isolation, Northern blotting and hybridization
Total ribonucleic acid (RNA) was isolated with TRIzol reagents according to the manufacturer’s instructions (Life Technologies, Karlsruhe, Germany) and aliquots of 10 µg were separated on 1·5% agarose/formaldehyde gels to visualize ribosomal bands. RNA was blotted onto Hybond-N+ nylon membranes (Amersham Pharmacia Biotech Europe, Freiburg, Germany) by capillary transfer. Equal loading and transfer of RNA was controlled by photography of the ethidium-bromide-stained gels before and after the transfer. The cDNA probes were labelled with [32P]-dCTP using the NonaPrimer Kit (Appligene, Heidelberg, Germany). The Expresshyb system was used for prehybridization and hybridization, as described by the manufacturer (Clontech, Heidelberg, Germany). Blots were autoradiographed and quantified using a PhosphorImaging System (Raytest, Straubenhardt, Germany). Poly(A)+ RNA was isolated from total RNA using the Oligotex mRNA kit (Qiagen, Hilden, Germany).
Suppression-subtractive hybridization was performed with a polymerase chain reaction (PCR) Select cDNA subtraction kit (Clontech). The method used two restriction enzyme-digested, adaptorligated cDNA pools. One contained specific transcripts (tester cDNA) and the other contained the reference transcripts (driver cDNA). After the subtraction of the tester cDNA pool from the driver cDNA pool, suppression PCR was carried out. This resulted in a selective amplification of transcripts more abundant in the tester pool (an excellent scheme of the SSH method is given by Diatchenko et al. 1996 and Gurskaya et al. 1996). The tester sample consisted of pooled mRNA isolated from leaf rosettes exposed to 20, 30 or 40 µm mercuric chloride. The concentration range was chosen according to previous results upon HgCl2 exposure using this ecotype (at a concentration of 20 µm, HgCl2 metallothionein 1 transcripts, known to be responsive to some metals, started to accumulate and morphological modifications of primary leaves became visible; 40 µm HgCl2 was the maximum concentration allowing growth; Heidenreich et al. 1999). The driver sample contained mRNA from untreated leaf rosettes. Because of demand in the later differential screening method, SSH was also performed exchanging tester and driver (reverse subtraction).
To verify the correct blunt-ended adaptor ligation of the RsaI digested tester and driver cDNA pools, primers for the Arabidopsis glycerol-3-aldehyde-phosphate gene GapC (M64116) were designed (5′-CAACGAGCACGAATACAAGTC-3′, 5′-GTCAACAACTGAGACATCAACG-3′). Using these gene-specific primers, a 328 bp cDNA fragment located between two RsaI restriction sites was amplified. Using these primers in combination with the adaptor primers, adaptor ligation efficiency was controlled. The PCR conditions were as described in the user’s manual.
Cloning and differential screening
The cDNA fragments obtained by the SSH method were cloned into a pGEM-T vector (Promega, Mannheim, Germany). Bacterial clones were stored at −80 °C as glycerol stocks for PCR amplification of the cDNA inserts.
For all further investigations, cDNA inserts of clones were amplified from 1 µL overnight culture using nested primers and PCR conditions as described in the manual of the PCR Select cDNA subtraction kit. All clones producing a single PCR-band were further investigated for differential expression using the Clontech PCR Select differential screening kit. Restriction digestion with AluI followed by-product separation on 3% agarose gels (Amersham Pharmacia) was used to analyse residual putative mercury-induced cDNA clones for duplicates. Plasmids were sequenced commercially (Qiagen) and sequences were analysed with programs of the Genetic Computer Group (Madison, WI, USA).
Production and use of cDNA arrays
The purified, PCR-generated cDNA fragments were adjusted to 20 ng µL−1 and denaturized by the addition of an equal volume of 0·6 N NaOH. One microlitre duplicates of each PCR product were spotted with an electronic pipette on Hybond N+ filters (186 spots per filter). After neutralization, washing and drying, the PCR products were fixed by UV cross-linking.
Poly(A)+ RNA (230 ng) was mixed with 2 µL random primers (Life Technologies) and denaturized at 70 °C for 10 min. After cooling on ice, 6 µL 5x first-strand buffer, 1 µL 100 mm DDT, 1·5 µL dNTP mixture (dATP, dTTP and dGTP, 20 mm each) and 10 µL α-[32P]-dCTP (111 TBq mmol−1) (Amersham Pharmacia) were added and samples were incubated for 10 min at 25 °C. For reverse transcription, 1·5 µL Superscript II reverse transcriptase (Life Technologies) were added and incubated for 90 min at 37 °C. The probes were purified with Quickspin columns (Roche Molecular Diagnostics, Mannheim, Germany).
Pre-hybridization was performed in 5 mL Easyhyb buffer (Clontech) containing 100 µg mL−1 denatured herring sperm DNA at 68 °C for 90 min. Hybridization was carried out in the same buffer, supplemented with 1 × 108 dpm of the freshly denatured probe, at 72 °C for 20 h. Membranes were washed at 72 °C, successively for 2 × 20 min in 2 × SSC, 1% SDS; 2 × 15 min in 0·5 × SSC, 1% SDS, and 2 × 15 min in 0·2 × SSC, 1% SDS. Filters were autoradiographed using a PhosphorImaging system (Raytest, Straubenhardt, Germany).
Results and discussion
Suppression subtractive hybridization
In this study, we examined the effects of transcript changes in Arabidopsis upon long-term HgCl2 treatment. The SSH procedure has been shown as a powerful method for the isolation of differentially expressed genes (Diatchenko et al. 1996; Gurskaya et al. 1996); to enrich differentially expressed, HgCl2-induced genes, this procedure was carried out (for details see Materials and methods). A flow chart depicting the procedure is given in Fig. 1. From SSH, 576 cDNA clones were obtained initially. Of these, 513 clones, showing a single band after PCR amplification, were investigated further by the PCR select differential screening method. Dot blots of these cDNA fragments were hybridized with radiolabelled forward- and reverse-subtracted cDNAs. According to the manufacturer, only clones revealing a forward/reverse hybridization quotient of ≥ 5 could be regarded as representing differentially expressed transcripts. A non-subtracted tester and driver, comparable with normal reverse Northern hybridization, were also used as probes. A tester/driver quotient of ≥ 2·5 was set for further analysis of the corresponding clones. Of the 513 clones, 123 cDNA fragments achieved both criteria; 156 cDNAs showed a quotient of ≥ 5 for subtracted cDNA probes; however, the quotient for non-subtracted probes was ≤ 2·5. Of these 156 cDNA clones, 80 fragments had a quotient of ≥ 10 for the subtracted probes and were investigated further. Therefore, 203 clones were analysed for duplicates by AluI fingerprinting. Fifty-two clones could be excluded and the residual 151 cDNA clones were sequenced. Sequence analysis revealed 17 ribosomal cDNAs and five mitochondrial-encoded cDNA clones, which were excluded from further analysis. The isolated ribosomal cDNAs indicated that a critical parameter is the poly(A)+ isolation, which has also been described by Sävenstrand et al. (2000). We excluded the mitochondrial cDNAs, as the cDNA array analysis was carried out with labelled reverse-transcribed poly(A)+ RNA (see Materials and methods). In addition, non-overlapping cDNAs detecting identical genes in databases were sorted out. Therefore, only one representative for each gene was kept. Finally, 90 clones were further analysed for HgCl2-induced transcript accumulation by Northern blotting.
Expression pattern of genes
For Northern blot analysis, seedlings were grown for 20–21 d in the absence or presence of 20, 30 or 40 µm HgCl2. Total RNA was isolated, starting at 0830–1000 h (control), 1100–1300 h (20 µm HgCl2), 1400–1530 h (30 µm HgCl2) and finished the next day at 0830–1000 h (40 µm HgCl2). The average size of the detected transcripts was 1·0–2·0 kb. The smallest was 460 b and the largest 4·7 kb. In this analysis, only changes in transcript abundance of greater than 1·5-fold compared with the controls in replicate experiments were used. Thirty-one putative HgCl2-induced cDNAs showed an increased transcript level at 30 µm HgCl2(Fig. 2; 12 clones are given as example). However, some of these showed a decrease in mRNA at 40 µm HgCl2 (Fig. 2). This might be caused by an inhibitory effect at higher HgCl2 concentrations. On the other hand, a diurnal rhythm might have existed, because the harvest extended over a period of 24 h. To investigate this in more detail, reverse Northern analysis was carried out. Clones were spotted in duplicate on macroarrays, which then were hybridized against radiolabelled cDNAs obtained from RNA isolated from soil-grown, untreated control plants harvested at 0900, 1200, 1500, 2100 and 0900 h, respectively. Because the most important input for the circadian rhythm of gene expression is light and (with minor impact) temperature changes (Barak et al. 2000), plant material grown on soil but under otherwise identical growth parameters could be used. The comparability of RNA isolated from agar- or soil-grown plants was demonstrated by identical diurnal amplitudes of diurnal-regulated transcripts, independent of growth on soil or agar (Heidenreich 1999). The reproducibility of identical spot intensities was proved by an in-pair comparison of the intensities, which gave a straight line. Similarly, individual labelling and hybridization experiments were proved to be reproducible (Heidenreich 1999). The time-course-dependent transcript level modulations for the above-mentioned 12 clones are given in Fig. 3. Diurnal rhythms were obvious for 10 clones. The amount of these mRNAs varied drastically between the day and night periods. Transcript levels started to increase from a stationary level at 0900 h and were highest at 1500–2100 h. During the night period, the levels declined, reaching the stationary level at 0900 h on the following day. For two clones, no diurnal modulation could be detected. By comparison of the Northern blot results with the macroarrays for these 12 clones, six cDNA clones turned out to be induced by HgCl2. Four clones were additionally under diurnal control. The remaining six clones from this set were only regulated diurnally. Running the same analysis for all 31 putative HgCl2-induced cDNA clones, seven remained to be induced exclusively by HgCl2 (Fig. 1), 14 were influenced by a diurnal circle and were additionally induced by HgCl2, while 10 genes were exclusively under diurnal control. This corresponds to 11% of the 90 clones analysed by Northern hybridization (14% when considering clones with multiple bands and without hybridization signal). This percentage calculation fits well to microarray analysis of diurnal-regulated genes in Arabidopsis (11% of 11 521 expressed sequence tags) (Schaffer et al. 2001). The published number of plant genes under the control of a circadian clock is still growing. The first examples of diurnally controlled gene regulation were mRNA transcripts of genes involved in photosynthesis and the early light-induced genes (Kloppstech 1985). Non-photosynthetic transcripts include catalases, wound-inducible, glycine-rich protein, RNA-binding protein, metallothioneins and Myb-related and transcription-factor mRNA (Heintzen et al. 1994; Green & Tobin 1999; Heidenreich et al. 1999; Somers 1999). In addition, microarray analysis of thousands of Arabidopsis genes resulted in an immense number of diurnal-regulated transcripts not involved in photosynthesis (Harmer et al. 2000; Schaffer et al. 2001).
The number of truly HgCl2-induced cDNAs corresponds to 14% of the 151 sequenced clones; this is in the percentage range of the recently published ozone-induced cDNAs, isolated also by the SSH method (Sävenstrand et al. 2000). Among the set of differentially expressed genes, 13 were expressed at levels 1·5–2·5 times higher in HgCl2-treated plants; six showed induction factors of 2·5–5·0; for two genes, the highest variation was an induction of between 5- and 10-fold (Table 1).
Table 1. Long-term HgCl2-induced genes in A. thaliana
Induction factors: + = 1·5–2·5; + + = 2·5–5·0; + + + = 5–10.aClones most probably identical to sequences in databases; bclones similar to sequences in databases.
Ath NADPH-protochlorophyllide-oxido-reductase B (PorB)a
The cDNA sequences were searched against the European Molecular Biology Laboratory (EMBL) database and the A. thaliana genome database(s). Table 1 lists the 21 clones, which were induced by long-term HgCl2 treatment. A high degree of matches to genomic sequences and annotations were obtained. Full-size gene products corresponding to isolated cDNA fragments could be identified and definitive or putative functions could be assigned.
One group of transcripts, induced by long-term mercury treatment, encoded proteins targeted to the chloroplast. Among these, the most affected was the PorB gene, which codes for the NADPH-protochlorophyllide-oxido-reductase B, a key enzyme in chlorophyll synthesis. Two other genes (1_f08 and 2_b02) encoding for light harvesting complexes were induced as well. There is evidence of a preferred interaction of Hg2+ with components of the photosynthetic apparatus – for example, the water-splitting complex, plastocyanin and NADPH-protochlorophyllide-oxido-reductase (Xyländer et al. 1996) – thus inhibiting photosynthetic electron transport and leading to oxidative damage. The induction of a putative chloroplast 50 S ribosomal protein L31 (3_b03) is an indication that the transcription machinery in the chloroplasts is also affected. The induction of these genes may help to compensate for damages caused by mercury.
In tomato seedlings exposed to mercury, a substantial increase in H2O2 was detected (Cho & Park 2000). This type of oxidative stress fits well to the elevated transcription of a plastidic copper/zinc superoxide dismutase (1_h01) together with a catalase (1_g07). The isolated catalase cDNA was identical to the Cat3 gene, which is highly expressed in leaves (McClung 1997). Three catalase genes are known in Arabidopsis (McClung 1997); the corresponding three gene products in different combinations are responsible for six tetrameric isoforms. The three genes are transcribed in an organ-specific manner, resulting in a corresponding distribution pattern of the isoforms. An increased catalase activity was found in roots and leaves of mercury-tolerant phenotypes of Chloris barbata compared with their non-tolerant counterparts (Patra, Lenka & Panda 1994). In tomato, the enzyme activities for superoxide dismutase, catalase and peroxidase were induced by mercury treatment (Cho & Park 2000). The jasmonate-inducible protein isolog (1_a10), found to be induced at the transcriptional level, might represent a link to lipid peroxidation. In tomato, mercury application resulted in the production of malondialdehyde, an indicator for lipid peroxidation (Cho & Park 2000).
Two transcripts (1_a04 and 1_f01), which were induced by Hg2+, encoded pEARLY1 and pectin methylesterase PME3, respectively. The two proteins are located in the cell wall. In a screening for aluminium-induced genes, pEARLY1 was isolated as a rapidly induced transcript for a secreted cell wall protein with a proline-rich middle stretch and a hydrophobic C-terminal (Richards & Gardner 1995). Interestingly, in our investigations only pEARLY1 transcripts accumulated at a concentration of 20 µm mercuric chloride in the medium; for all other transcripts, elevated steady-state levels were only detected at 30 µm mercuric chloride or even higher.
Two transcripts (1_b03 and 2_e01) encoding cytochrome P450 proteins with unknown function were also induced by mercuric chloride. P450 enzymes are involved in the biosynthesis of a variety of secondary metabolites (Bolwell, Bozak & Zimmerlin 1994), and an accumulation of anthocyanins upon mercury treatment was found in several Arabidopsis ecotypes (Heidenreich 1999).
Another group of transcripts encoded proteins with supposed functions in signalling and transcriptional regulation (1_b03, 1_b06 and 3_b05), but the specific functions are hitherto unclear. For other members of this group (1_c02, 2_e02), only annotations as putative proteins were found in the databases. Nevertheless, the domain analysis suggested a role in signalling. The Trp-Asp repeats signature characteristic for G-beta proteins, important in signal transduction, is found in both genes. The indole-3-glycerol phosphate synthase (1_a07) is involved in the biosynthetic pathway leading to tryptophan, from which a variety of secondary metabolites such as indolic acids and the phytoalexin camalexin are formed. It was shown that oxidative stress, as well as abiotic elicitor treatment, resulted in the synthesis of camalexin (Zhao, Williams & Last 1998). The function of mercuric chloride as an abiotic elicitor of phytoalexin biosynthesis is also well known (Moesta & Grisebach 1980). However, the induction of the biosynthetic pathway for tryptophan and the derived secondary metabolites in Arabidopsis by mercuric ions has so far not been investigated.
In summary, several of the genes described above are also induced upon other stress conditions. This clearly points to a cross-talk between different induced defence responses and to multiple pathways in transducing these responses, as described for pathogen attack (Maleck & Dietrich 1999). A complex signalling network that involves many factors will finally contribute to a subtly modulated form of gene expression upon various stimuli. We didn’t find induced genes involved in heavy metal detoxification (e.g. phytochelatin or metallothionein production). However, phytochelatin synthase appears to be expressed independently of heavy metal exposure (Cobbett 2000) and Cd2+-induced transcript changes of a recently isolated cDNA from wheat involved in phytochelatin synthesis could only be demonstrated by the highly sensitive RT-PCR (Clemens et al. 1999). Metallothioneins have also been shown to be expressed constitutively in Arabidopsis leaves, having only a slight increase in MT1a abundance upon HgCl2 treatment (Heidenreich et al. 1999). Interestingly, 10 clones were also under diurnal control (Fig. 3, Table 1). This included transcripts involved in photosynthesis (Lhb, PorB) and the antioxidative pathway (CAT3), as already reported in the literature (Somers 1999). The diurnal transcript profile of the other genes suggested that an endogenous clock might control additional genes. A similar conclusion was reached by microarray profiling of differentially expressed genes, including several abiotic-stress-induced transcripts, in dark- and light-grown Arabidopsis (Desprez et al. 1998; Harmer et al. 2000; Schaffer et al. 2001).
The present study identified a number of marker genes for Hg2+ and it reveals a strong modulation effect of diurnal regulation. In further studies, these genes should be analysed upon short-term heavy metal treatment. Are there great differences with long-term HgCl2 application? The impact of other abiotic/biotic stresses should also be compared, as has been described for a few genes upon ozone fumigation and UV-B irradiation (Heidenreich et al. 1999; Sävenstrand et al. 2000). This will finally contribute to a better understanding of cross-talk pathways and gene expression, influenced by multiple environmental changes.
We wish to thank our colleague J. Durner for reading the manuscript critically. Seeds of A. thaliana were kindly provided by Czaba Koncz (MPI für Züchtungsforschung, Köln, Germany).