5S rRNA genes expression is not inhibited by DNA methylation in Arabidopsis

Authors


For correspondence (fax +33 473407777; e-mail Sylvette.Tourmente@geem.univ-bpclermont.fr).

Summary

Methylation has often been correlated with transcriptional inhibition of genes transcribed by polymerase II, but its role on polymerase III genes is less well understood. Using the genomic sequencing technique, we have analysed the methylation pattern of the different 5S-rDNA arrays of the Arabidopsis genome. Every cytosine position within the 5S sequence is highly methylated whatever the context – CpG, CpNpG or non-symmetrical. The methylation pattern of both transcribed and non-transcribed 5S units is similar, with no preferential methylated or unmethylated site. These results, taken together with 5-azacytidine treatments and in vitro transcription experiments using methylated 5S templates, demonstrate that 5S rRNA gene transcription is not inhibited by methylation. Non-transcribed 5S arrays are more subject to transition mutations resulting from deamination of 5-methylcytosines, leading to CpG depletions and an increasing A + T content. As there were no detectable differences in methylation, this implies more efficient repair and/or selection pressure in transcribed 5S-blocks.

Introduction

Arabidopsis thaliana accession Columbia is known to contain approximately 1000 copies of 5S rRNA genes per haploid genome. These occur in tandem arrays localized in the pericentromeric heterochromatin of chromosomes 3, 4 and 5 (Campell et al., 1992; Fransz et al., 1998; Murata et al., 1997). Two loci are present on chromosome 5: a minor one on the right arm (Kotani et al., 1999) and a major one on the left arm (Tutois et al., 1999). Three loci are present on chromosome 3 (loci 1, 2, 3; AGI, 2000; Cloix et al., 2001). Each 5S rDNA unit consists of a 120 bp transcribed sequence and an approximately 380 bp non-transcribed spacer, and can be recognized by a specific DNA-sequence signature (a T stretch) 3′ of the transcribed sequence (Cloix et al., 2001). In Arabidopsis thaliana both a major (85%) and some minor 5S RNA transcripts are produced. The 5S rDNA units homologous to major and minor (differing from the major one by one or two bases) 5S transcripts have been identified and belong to chromosome 4 or the chromosome 5 major block (Cloix et al., 2001).

In eukaryotic genomes, methylation of cytosine residues, 5-methylcytosines (5mC), commonly occurs in repetitive sequences at higher levels than the rest of the genome (Kovarik et al., 2000; Martinez-Zapater et al., 1986; Pruitt and Meyerowitz, 1986; Vongs et al., 1993). Methylation of repetitive sequences might suppress recombination between repeats in different genomic positions which otherwise would lead to translocations and other chromosomal rearrangements (Bender, 1998). Methylation can also cause a loss of RNA-polymerase-II-dependent transcription of repetitive sequences such as transposable elements, either by preventing transcription initiation or by impeding transcription elongation (Bender, 1998).

Numerous studies have shown a correlation between promoter methylation and the inactivation of genes transcribed by RNA polymerase II. The role of methylation in transcription of RNA polymerase I and III-dependent genes is less well understood. 5S genes, together with tRNA, 7SL RNA, SINEs, U6 snRNA and a few other small stable RNAs, are transcribed by polymerase III (pol III). The impact of DNA methylation on pol III transcription is not clear and relies on a small number of reports, all of which (except one in Brassica) use animal systems. Most of these studies were done on mammalian SINEs (Liu and Schmid, 1993), viral genes (Juttermann et al., 1991) and a tRNALys gene (Besser et al., 1990), and report a more-or-less severe transcriptional inhibition following methylation. In Brassica, using an in vitro assay, Arnaud et al. (2001) reported a transcriptional inhibition that is directly correlated to the level of the template methylation (a Brassica SINE). While the general tendency for methylation is to inhibit pol III transcription, exceptions exist. For example, in vitro methylation of somatic 5S rDNA from Xenopus laevis does not influence its transcription in Xenopus oocytes (Besser et al., 1990).

So far, analysis of methylation patterns of plant 5S rRNA genes has been limited to the use of methylation-sensitive restriction endonucleases, except in tobacco where genomic sequencing has been realized (Fulnecek et al., 1998). Using restriction enzymes, 5S rDNA has been shown to be highly methylated in different plant species: maize (Mascia et al., 1981); wheat (Grellet and Penon, 1984); lupin (Rafalski et al., 1982); Arabidopsis thaliana (Finnegan et al., 1996; Vongs et al., 1993); pea (Ellis et al., 1988); and flax (Goldsbrough et al., 1982). In tobacco, genomic sequencing of 5S rRNA genes using the bisulfite method showed that almost any cytosine residue might be methylated (Fulnecek et al., 1998). These authors did not find any non-methylated 5S units, and concluded that methylation of only some critical positions, or alternatively the density of 5mC, may be involved in the control of transcription.

In this paper we analyse the distribution of 5mC in 5S rDNA units of different 5S arrays of the Arabidopsis thaliana genome. The methylation patterns of 5S units homologous to 5S RNA compared to non-transcribed 5S units was established, as well as the impact of methylation on 5S transcription. Furthermore, an analysis of the mutations has revealed that non-transcribed 5S arrays are more subject to transition mutations resulting in an increasing A + T content.

Results

Evaluation of the methylation level of 5S genes by genomic sequencing

In order to elucidate the pattern of methylation along the 5S rDNA units, we performed bisulfite analysis of genomic DNA extracted from 14-day-old-rosettes. At this stage, as well as at 2, 4, 6, 10 and 16 days, and in leaves, flowers, stems and siliques, 85% of the 5S transcripts correspond to the major 5S RNA and 15% to minor ones (Cloix et al., 2001).

The bisulfite reaction converts non-methylated cytosines in DNA to uracils, while leaving 5mC unaltered (Clark et al., 1994; see Experimental procedures). We previously defined a sequence signature specific for each 5S block (a T stretch 3′ from the transcribed region) included in the genomic sequencing PCR product and defining the 5S array origin (Cloix et al., 2001). After bisulfite treatment, the two strands are no longer fully complementary and primers allowing specific amplification of the plus and minus strand can be designed. As methylation levels and number of cytosine residues in each context were similar on the two strands, we report only minus strand results. The sequences of the primers [5S-7(-) and 5S-8(-)] are given under Experimental procedures. They were designed with two imperatives: they should amplify 5S units from the different 5S arrays of the genome, and this should be independent of the methylation status of the cytosines. The PCR products were analysed over 252 bp containing 60 bp of 5′ flanking region, the 120 bp transcribed region and 72 bp of the 3′ region containing the T-stretch block-specific signature.

We first determined the overall proportion of 5mC in 5S units from the different 5S arrays (Table 1). We did not obtain 5S units from the small 5S arrays (locus 3 of chromosome 3 and minor block of chromosome 5) as they represent only a low fraction of the 5S units of the genome (5% each; Cloix et al., 2001). The methylation level, in percentage of methylated cytosines, is very high in all the 5S arrays: 79.3% on chromosome 4 (26 sequences), 74.8% in chromosome 5 major block (32 sequences), and 83.3% in chromosome 3 loci 1 and 2 (30 sequences). We cannot clearly distinguish the chromosome 3 loci 1 and 2, as the T stretch differs by only one base (of about 20 residues; Cloix et al., 2001). This general ratio of 5mC to total cytosines in 5S rDNA units (79.1%) has to be compared to the low general level of methylation (6.3%) of the Arabidopsis thaliana genome (Meyerowitz, 1992; Wagner and Capesius, 1981). The only comparable results available are in Nicotiana tabacum, where a 50% methylation of 5S units is observed in a general level of methylation of 30% for the whole genome (Fulnecek et al., 1998).

Table 1.  Methylation frequencies (%) of C residues in 5S rDNA units from different 5S arrays of the Arabidopsis genome
 Entire
unit
5′ spacerTranscribed
region
3′ spacer
  1. 5′spacer, transcribed region and 3 ′spacer are given in Figure 1.

Chromosome 479.392.88061.6
Chromosome 5  major block74.878.67568.6
Chromosome 3  loci 1, 283.384.783.381.6
Mean value79.185.379.470.7

The pattern of heavy methylation observed for the 5S genes of Arabidopsis was similar for the different 5S arrays of the genome. This comparable level of methylation between the different arrays was also observed for the different regions of the 5S units, with a decreasing level from spacer 5′ to spacer 3′ (the lowest mean value still being 70.7%; Table 1).

Identification of cytosine methylation sites in 5S rDNA units

The precise sequence context of 5S methylation was studied further. As there are no significant differences in methylation levels between different 5S arrays (see above), all arrays were considered together. As shown in Table 2, 5S units' cytosines in CpG context were methylated at the level of 93.4% (95.9% in spacer 5′, 92.7% in transcribed region, 92.8% in spacer 3′). Cytosines in CpNpG were methylated at 85.6% (87.5% in spacer 5′, 85.3% in transcribed region). The more surprising results came from cytosines in non-symmetrical sites, with 72.6% methylation, usually significantly less methylated than symmetrical sites. These are the highest values reported to date for transcribed sequences. We did not observe a preferential methylation on the non-symmetrical site CpA/TpA, as reported for Brassica napus SINE S1 (Goubely et al., 1999); this was also not observed for Nicotiana tabacum 5S units (Fulnecek et al., 1998).

Table 2.  Methylation frequencies of C residues in the different contexts
 5′ spacerTranscribed
region
3′ spacerMean
CpG95.992.792.893.4
CpNpG87.585.3no site85.6
Non-symmetrical80.772.465.672.6

Evaluation of the methylation status of 5S genes in relation to their transcription

In a preceding work (Cloix et al., 2001), using a combination of 5S DNA and RNA sequences, we were able to identify 5S units identical to 5S RNAs as well as 5S units non-identical to 5S RNAs. Chromosome 5 minor block and chromosome 3 loci 1, 2, 3 are considered to be silent, as we never found 5S RNA identical to the 5S units present in these blocks. These 5S units contain numerous mutations when compared to the 5S RNA sequence. They are not contributing to the mature 5S RNA pool, and are considered as silent or not transcribed in this paper, although we cannot exclude a low transcription rate followed by a rapid degradation of these 5S RNAs. Transcribed units are present only on chromosome 4 and the chromosome 5 major block. Eighty-five per cent of the 5S RNAs correspond to the major RNA, and the other 15% are minor RNAs differing from the major one by one or two bases substitutions.

The methylation level of each cytosine in 5S units, identical or not to 5S RNAs, is given for 5S units from chromosome 5 major block in Figure 1, according to their position in the spacer or the transcribed region. In Arabidopsis, the internal promoter has been defined by sequence homology with Xenopus sequences (Bogenhagen et al., 1980). Box A, intermediate element (IE) and box C, which are sites for the fixation of transcription factors, are present in the transcribed region of the 5S units. On the basis of sequence comparisons (Venkateswarlu et al., 1991), some elements of the 5′ flanking region are also considered to be important for 5S transcription: a TATA sequence; a GC dinucleotide; and a universally conserved C nucleotide, one base before the initiation site. Using in vitro transcription experiments, we have shown that the TATA motif is necessary for transcription (Cloix et al., 2001).

Figure 1.

Distribution of cytosine methylation on Arabidopsis 5S rRNA genes from chromosome 5 major block.

The sequence is orientated 3′−5′ because it is the minus strand. The transcribed sequence is indicated in brackets; grey boxes show promoter elements. Sequences labelled with an asterisk are homologous to the major 5S RNA. The consensus sequence (CONS) was established using all the 5S units of the chromosome 5 major block. Sequence variations between PCR clones and the consensus sequence are indicated. Dashes indicate nucleotide deletions. Clone sequences are followed by the percentage of cytosines methylated. Framed numbers are the mean methylation value for each type of unit. Cytosine methylation at ●, symmetrical CpG; ▪, symmetrical CpNpG; ▴, non-symmetrical C residue; non-methylated C residues at ○, symmetrical CpG; □, CpNpG; ▵, non-symmetrical positions.

As shown in Figure 1, there is no characteristic methylation difference between the two types of units (78.8 and 73.4%, respectively). No preferential methylation site was observed in promoter sequences. Unmethylated sites appear to be randomly distributed, with no preferential unmethylated site which could be related to transcription as observed for the 18S-25S-5.8S rDNA in other species (cereal hybrids, Houchins et al., 1997; wheat, Flavell et al., 1988; Sardana et al., 1993; maize hybrid, Jupe and Zimmer, 1993; pea, Kaufman et al., 1987). In agreement with our data, Chen et al. (1998) did not observe such a site in Arabidopsis and Brassica, nor did Fulnecek et al. (1998) find any particular unmethylated site in Nicotiana tabacum 5S rDNA units.

The major 5S RNA accounts for 85% of the 5S RNAs, thus representing a major fraction of the cell transcripts. We have estimated the number of 5S rDNA units homologous to the major 5S RNA to be only 80 copies, present uniquely on chromosome 4 and the chromosome 5 major block (Cloix et al., 2001). Thus these 5S units (or at least some of them) should be active in transcription. We have analysed the methylation status of 20 of these units. Those found on the chromosome 5 major block are presented in Figure 1 (labelled with an asterisk). They are all highly methylated, with a mean value of 78%, comparable to the mean value of 73% obtained for non-transcribed units from chromosome 4 which contain a TAGA motif (instead of the TATA motif) inhibiting transcription, but having a transcribed region homologous to the major 5S RNA.

In addition, using primers selected to amplify only low-methylated 5S rDNA units (5S-BIAIS-3 and 5S-BIAIS-4; see Experimental procedures), we were able to amplify 10 5S units with levels of methylation ranging from 12 to 38%. These 5S units belong to all 5S arrays, transcribed (chromosome 4, chromosome 5 major block) or not (loci 2 and 3 of chromosome 3), but do not correspond preferentially to 5S RNA (only 1 for 10 sequences).

Impact of methylation on in vitro transcription

The impact of DNA methylation on the in vitro transcription of a 5S rDNA unit was tested. A construct containing a 5S rDNA gene corresponding to the major 5S RNA (Cloix et al., 2001) was methylated using the HpaII methylase (not shown) or the CpG-specific SssI methylase (Figure 2; see Experimental procedures). Three methylated samples with increasing methylation levels (34.6, 55 and 97%), as well as the non-methylated control, were used for in vitro transcription; methylation levels were determined using HpaII digestion. As shown in Figure 2, we observed a similar quantitative transcription for all these samples, even for the most methylated one, indicating no inhibition of in vitro transcription by methylation.

Figure 2.

In vitro transcription from methylated 5S gene with tobacco nuclear extract.

The 1037 construct, containing a complete 5S gene, was methylated with the CpG-specific SssI methylase. The reaction was repeated two and three times after the first treatment by adding fresh enzyme, and aliquots were taken after each methylation treatment (1, 2 and 3). Thus samples 1, 2 and 3 correspond to the 1037 clone with increasing methylation levels (34.6, 55 and 97%, respectively); control represents a non-methylated 1037 clone.

Effect of 5-azacytidine on 5S transcription

To test in vivo the role of cytosine methylation on 5S gene transcription, Arabidopsis seeds were germinated in the presence of 0.3, 0.5 and 1 mm 5-azacytidine (5-azaC), a cytosine analogue that inhibits cytosine methylation (Haaf, 1995; Jones, 1985). Germination in the presence of this drug leads to a dwarf phenotype, the gravity of which increases with drug concentration (Figure 3a). The effect of 5-azaC treatments on 5S rDNA methylation was estimated by digesting DNA from control and treated plants with HpaII followed by hybridization with a 5S probe and quantification of the 0.5 kb band. The intensity of this band increases in the treated plants proportionally to 5-azaC concentration, up to 15 times more than in the control with 1 mm 5-azaC (Figure 3b). Thus the treatment with 5-azaC increased the accessibility of Arabidopsis 5S genes to HpaII, showing a decrease in 5S rDNA methylation.

Figure 3.

Effect of 5-azacytidine treatment on 5S rRNA gene expression.

(a) Phenotype of 14-day-old plantlets germinated either in water (control) or in 5-azaC at the concentrations indicated. Bar = 5 mm.

(b) Southern analysis of control and 5-azaC-treated plant genomic DNA cleaved with HpaII and probed with a 5S probe. The demethylation induced by 5-azaC was estimated by monitoring the intensity of the 0.5 kb band (5S unit size) compared to the rest of the DNA by phosphorimager scanning for each digestion.

(c) Northern analysis of control and 5-azaC-treated plants. Total RNA (300 ng) was hybridized with a 5S probe. Phosphorimaging quantification of the 5S signal is indicated in the lower part of the panel (black values). The 25S band intensity (values indicated in white) on the ethidium bromide (EtBr) stained gel was used as loading control to normalize the 5S signals to a same amount of total RNA. Normalized 5S signal intensities are indicated by the ratio 5S/25S as well as the decreasing factor (1/ratio 5S/25S) of the 5S signal in treated plants compared to control.

We monitored 5S RNA expression levels in 5-azaC-treated plants versus untreated control plants by Northern blot analysis of 14-day-old plant RNA (Figure 3c). The amount of total RNA spotted per lane was estimated by quantifying the 25S band intensity on the ethidium bromide-stained profile, and the 5S signals obtained after hybridization were normalized according to these values. As shown in Figure 3(c), the 5S signal level was decreased 1.06, 1.39 and 1.52 times in the treated plants for 5-azaC concentrations of 0.3, 0.5 and 1 mm, respectively. However, according to Chen et al. (1998), polymerase I transcription (18S, 5.8S and 25S) is increased approximately twofold in treated Arabidopsis suecica plants. Thus transcripts whose transcription level is stable appear about twofold less represented in total RNA from treated plants than in the same quantity of total RNA from the control. The 5S transcription level is therefore probably not affected by 5-azaC treatment, and the diminution observed probably reflects the polymerase I transcription upregulation.

Analysis of mutations in the different 5S arrays

5mC residues are often hot-spots for spontaneous transition mutations, as spontaneous deamination of 5mC occurs frequently in DNA, leading to thymine. Arabidopsis 5S rDNA is highly methylated, and we have looked for the proportion of the different types of mutations (Table 3). C→T and G→A transitions were expected to be more abundant than the other mutations, according to this argument.

Table 3.  Mutation frequencies in the different 5S arrays
BlocksChromosome%A + TTransitionsN→(A or T)N→(C or G)
C→T & G→AT→C & A→G
  1. (N = A, T, C or G). Calculations were made on amplification products obtained from YACs. So loci 2 and 3 of chromosome 3, which are present on different YACs can be considered separately. rs, reference sequence.

Transcribed5 major block56.8rsrsrsrs
457.516.923.86634
Silent5 minor block59.23248119
3 locus 26237.615.27129
3 locus 161.6347.579.720.3

A small proportion of 5S units from chromosome 5 major block (22%) has the potential to be transcribed in the major 5S RNA (Cloix et al., 2001). We established a consensus sequence from these 5S units, and used it as the 5S sequence reference to evaluate the mutations present in the different 5S arrays of the genome. 5S DNA sequences homologous to major or minor 5S RNAs are present mostly in the chromosome 5 major block, but also in chromosome 4 to a lesser extent. None was found in chromosome 3 (loci 1, 2, 3) or chromosome 5 minor block, which are considered as silent (Cloix et al., 2001).

Twenty 5S sequences from chromosome 4, nine from chromosome 5 minor block, seven from chromosome 3 locus 2, and 14 from chromosome 3 locus 1 were aligned with the consensus sequence reference. We first evaluated the A + T percentage for each 5S array of the genome. Transcribed blocks have an A + T content of 57%. Silent blocks contain significantly higher A + T contents (61%; P = 0.05, Z-test). We also evaluated the proportion of the different mutation types, considering the 5S-arrays as two groups. The first group includes the silent 5S arrays. The highest mutation values are for C→T and G→A transitions, presumably resulting from deamination of 5mC (for comparison, the values for the inverse mutations T→C and A→G are given). The total of mutations resulting in A or T is between 71 and 81% and explains the relative A + T richness of these 5S sequences. The second group includes 5S arrays containing transcribed 5S units (chromosome 5 major block and chromosome 4). In chromosome 4 (we do not have values for chromosome 5 major block, which is the reference), the proportions of C→T and G→A deamination mutations are comparable to those of the inverse transition mutations (T→C and A→G), and mutations leading to A or T are less frequent than in non-transcribed blocks (66%; P = 0.05, Z-test).

CpG and CpNpG depletion

5mC deamination has been shown to result in CpG depletions within or in the immediate vicinity of the Arabidopsis genes analysed, although there was no CpNpG depletion (Gardiner-Garden et al., 1992). To date, no data have been available upon CpG and CpNpG depletions in the highly methylated fraction, and we have analysed both CpG and CpNpG depletion in the different 5S arrays of the Arabidopsis genome. The results are presented in Table 4. The transcribed-consensus sequence from chromosome 5 major block, used as reference, contains 15 CpG and 11 CpNpG motifs. The chromosome 4 array, which contains 5S transcribed units, contains a mean of 93% of the CpG, whereas untranscribed blocks (chromosome 3 loci 1, 2, 3) have lost more CpG motifs. In contrast, there was no significant depletion of CpNpG motifs, in agreement with the results obtained with single or low-copy genes by Gardiner-Garden et al. (1992).

Table 4.  CpG and CpNpG depletions in 5S arrays
ChromosomeCpG (%)CpNpG (%)
  1. The reference (from chromosome 5 major block) contains 15 CpG and 11 CpNpG motifs, representing 100%. Calculations were made on amplification products obtained from YACs, so loci 2 and 3 of chromosome 3 can be considered separately.

5 major block15 (100)11 (100)
414 (93)10 (93)
3 locus 312 (86)11 (100)
3 locus 29 (60)9 (82)
3 locus 19 (60)10 (91)

Discussion

We have observed that 5S rRNA genes have an average level of 5mC of 79%. This is very high compared to the low level of methylation for the Arabidopsis genome (6.3%; Meyerowitz, 1992; Wagner and Capesius, 1981). In Nicotiana tabacum, 5S rDNA units have an average level of 5mC of 50%, with an average of 30% observed over the whole tobacco genome (Montero et al., 1992). The level of 5mC in plant DNA has been suggested to be related to genome sizes, and the Arabidopsis genome is 20 times smaller than the Nicotiana genome (Jeddeloh and Richards, 1996; Kovarik et al., 1997). In this context, the higher level of methylation of Arabidopsis 5S rDNA units (79% compared to 50% in Nicotiana) is surprising. Both species contain a similar quantity of 5S rDNA units (1000), all situated in heterochromatic regions (Kovarik et al., 2000). In Arabidopsis thaliana, all the 5S arrays studied harbour the same level of methylation, and all are situated in pericentromeric regions (Fransz et al., 1998; Murata et al., 1997). Nevertheless, we found some less-methylated 5S units from both transcribed or non-transcribed 5S blocks, showing that there is a methylation heterogeneity in repetitive DNA of high sequence homogeneity.

A survey of methylation patterns in Arabidopsis indicates that highly repetitive sequences, such as 18S-5.8S-25S rDNA arrays and centromere-associated repeated sequences, are densely methylated (Vongs et al., 1993). Methylation correlates with reduced homologous recombination between methylated regions, and therefore methylation of repetitive sequences might suppress recombination between repeats in different genomic position (for review see Bender, 1998), as for 5S arrays in the Arabidopsis genome. Repetitive sequences such as transposable elements are usually inactive when positioned in heterochromatic regions. The 18S-25S rDNA, although heterochromatic and transcribed, is less methylated than 5S rDNA (Kovarik et al., 2000). 5S is an exception in that it is an active genetic region that remains highly methylated, and 5S units represent the most heavily methylated transcribed sequences.

Symmetrical CpG and CpNpG sites have been demonstrated as the most frequent target for cytosine modification in plant DNA (Gruenbaum et al., 1981). In Arabidopsis 5S rDNA, cytosines in CpG are more methylated (93.4%) than cytosines in CpNpG (85.8%), a situation also observed for tobacco 5S rDNA as well as other heterochromatic sequences (Kovarik et al., 2000). According to Kovarik et al. (1997) in Arabidopsis, the CpNpG methylation appeared to be highly compartmentalized into the repetitive fraction and limited in plant promoter sequences of single- or low-copy genes. Methylation at non-symmetrical cytosines has been described in mammals (Tasheva and Roufa, 1994), fungi (Selker et al., 1993) and plants (Fulnecek et al., 1998; Goubely et al., 1999; Ingelbrecht et al., 1994; Meyer et al., 1994; Wang et al., 1996), but is less present than methylation at symmetrical sites. The 72% level of non-symmetrical methylation for Arabidopsis 5S rDNA is noteworthy. It is much lower in tobacco 5S rDNA (14%), resulting in a global lower methylation in Nicotiana 5S rDNA (50% compared to 79% in Arabidopsis).

We detected no significant difference in methylation patterns between Arabidopsis 5S units identical, or not identical, to 5S RNAs. No preferential unmethylated site was observed, in agreement with Nicotiana 5S results (Fulnecek et al., 1998).

It is now clear that a high degree of methylation of the 5S gene is compatible with transcription by RNA polymerase III. From our results it appears unlikely that the density of 5mC is involved in the control of transcription for several reasons: (i) there is no quantitative or qualitative difference in the methylation pattern between 5S rDNA genes identical or not identical to 5S RNAs; and (ii) the 5S transcription is not affected by 5-azaC-induced decreased methylation. Chen and Pikaard (1997b) and Chen et al. (1998) have shown that polymerase I transcription was stimulated by reduced cytosine methylation in Arabidopsis suecica and Brassica. This has been studied in the nucleolar dominance phenomenon, where one parental set of ribosomal 45S rRNA genes is transcribed and the other one silent in plant hybrids. Demethylation of either 45S rRNA genes or of a regulatory locus distinct from these genes causes derepression of the silent (or underdominant) genes (Pikaard, 1999). Possibly, the same repression mechanisms that control the nuclear dominance in hybrids may be responsible for the control of the number of active 45S rRNA genes within a pure species (Wallace and Langridge, 1971), and probably reflect a dosage-compensation mechanism (Pikaard, 1999). The control of active 5S rDNA genes, independent of methylation status, could be different. Using in vitro transcription we have shown that some mutations, largely represented in 5S units, abolish transcription and could limit the number of active genes (Cloix et al., 2001).

5S rDNA in Arabidopsis is methylated at a higher level compared to 18S-25S (Kovarik et al., 1997), and both methylation and condensation of 5S rDNA are higher than for 18S-25S rDNA in tobacco (Kovarik et al., 2000). Similarly, 5S rDNA in Brassica appeared to be more resistant to hypomethylation (Chen and Pikaard, 1997a). Transcription of 5S and 18S-25S rDNA by polymerase III and polymerase I, respectively, might be unequally sensitive to cytosine methylation density. Finally (iii) the three different methylated 5S samples gave identical in vitro transcription results with tobacco nuclear extract.

Using the same in vitro transcription system, Arnaud et al. (2001) have shown an inhibition of transcription from Brassica SINE S1 sequence proportional to the methylation level of the templates, with complete inhibition using the highly methylated template. Thus 5S genes and retroposons SINEs, both using polymerase III for their transcription, are differently regulated by methylation. It could be postulated that transcription regulation mechanisms mediated by methylation differ between 5S rRNA genes and SINEs because of the impact of their transcription on the genome. Indeed, SINEs transcription could be deleterious for the genome because it could be at the origin of a large spectrum of mutations and genome reorganization. The genes transcribed by polymerase III represent a functionally diverse group, and inside this group 5S and SINEs are, respectively, type I and type II genes, differing in their promoter elements (Willis, 1993). Methylation appears to have a different impact on polymerase III genes transcription according to the nature of their promoter sequences. The plant results with Brassica SINE S1 and Arabidopsis 5S rRNA genes are in agreement with mammalian reports, where SINEs (Liu and Schmid, 1993), viral genes (Juttermann et al., 1991) and tRNALys gene transcription (Besser et al., 1990) were inhibited by methylation, although the transcription of a Xenopus 5S rRNA gene was not (Besser et al., 1990).

Spontaneous deamination of 5mC occurs frequently in DNA. Despite the existence of a mismatch-repair system, 5mC acts as an intermediate in C→T mutagenesis (Coulondre et al., 1978; Gardiner-Garden et al., 1992; Kricker et al., 1992; Radman and Wagner, 1986; Rideout et al., 1990; Selker, 1990). We found high levels of C→T transitions in 5S rDNA of Arabidopsis, particularly in non-transcribed arrays. These C→T transitions may largely explain the A + T richness of non-transcribed 5S arrays relative to transcribed 5S sequences. Transcribed 5S sequences have an A + T content (57%) similar to that given for coding regions (56%), and non-transcribed 5S units have an A + T content (61%) lower than that given for non-coding regions (68%; AGI, 2000), but identical to the A-T content of centromeric sequences (61%), tandemly repeated or dispersed and non-coding (Tutois et al., 1999). Gardiner-Garden et al. (1992) reported elevated levels of TpG and CpA, the potential deamination mutation products of methylated CpG, with a CpG deficiency in Arabidopsis genes, but no CpNpG depletion in relation to 5mC deamination events. But according to Kovarik et al. (1997), the CpNpG methylation would be limited in promoter sequences of single- or low-copy genes. We found CpG depletions in Arabidopsis non-transcribed 5S arrays, and a lower CpNpG depletion in relation to 5mC deamination events, even taking into account the lower methylation level of cytosines in CpNpG compared to CpG. So for classical genes, as for highly methylated 5S genes, methylation does not lead to CpNpG depletion, and CpNpG depletion may not be a characteristic of angiosperm genes. This lower CpNpG depletion could imply the existence of a CpNpG mismatch-repair system more accurate and efficient than that for CpG. We have shown that non-transcribed 5S arrays mutate more often than transcribed ones. Sequences of rDNA repeats are believed to be maintained constant by genetic recombination and gene-conversion events (Flavell, 1985; Klein and Petes, 1981; Morzycka-Wroblewska et al., 1985). There is thus a selective pressure and/or more efficient mismatch-repair mechanism in transcribed compared to non-transcribed 5S arrays. CpG depletion of the Arabidopsis gene sequences is similar to that of gene sequences from other angiosperm species, notwithstanding the lesser methylation of Arabidopsis relative to other angiosperms. Although a similar proportion of total CpG is methylated in angiosperms and vertebrates, angiosperm genomes are not as CpG-depleted as those of vertebrates, and this could arise from different efficiencies of repair (Gardiner-Garden et al., 1992).

In conclusion, all 5S arrays, transcribed or not, are similarly methylated, and the same observations have been made at the level of the 5S genes. Moreover, methylated templates were found fully active for in vitro transcription. These results suggest that methylation does not impair transcription-factor binding on 5S rDNA, neither in vitro nor in vivo, and that methylation does not regulate transcription of 5S rRNA genes. Methylation on 5S gene clusters and on neighbouring repeated non-coding sequences (such as the 180 bp satellite) could have the same function, i.e. suppression of the recombination between repeats. When looking for 5S transcription, we did not find 5S RNAs homologous to some 5S genes, although these have the potential to be transcribed (Cloix et al., 2001). Thus either the transcripts of these genes are unstable and degraded, or some 5S genes from the transcribed blocks are silent.

Experimental procedures

Plant material and 5-azacytidine treatment

Arabidopsis thaliana accession Columbia was used. For 5-azacytidine (5-azaC) treatments, seeds were surface-sterilized and germinated for 3 days at 25°C in Petri dishes containing either water or a solution of 5-azaC (0.3, 0.5 or 1 mm prepared fresh every day; Sigma, St Quentin Fallavier, France). After germination, seeds were grown on a germination medium [MS Salt (Sigma) supplemented with 3% sucrose and 0.8% Bacto-Agar] for 14 days in a growth chamber using a 16 h light/8 h dark growth regime.

Nucleic acids isolation and gel-blot analysis

Total genomic DNA was isolated from rosettes using the CTAB method (Doyle and Doyle, 1987). The digestions were realized with 250 ng of Arabidopsis genomic DNA and 15–20 units of restriction enzyme in the recommended buffer (New England Biolabs, St Quentin, Yvelines, France and Roche, Moylan, France), supplemented with 1 × BSA. Digested DNA was electrophoresed in 0.8% agarose gels overnight, depurinated in 0.25 N HCl and capillary blotted to hybond-N+ membranes (Amersham, Orsay, France).

Total RNA was extracted using the method described by Logemann et al. (1987) with minor modifications. Briefly, plant tissue (100 mg) was ground in a tube with 400 µl guanidine buffer (8 m guanidine hydrochloride, 20 mm MES, 20 mm EDTA and 50 mm 2-mercaptoethanol pH 7.0). Proteins were removed by two successive extractions with phenol : chloroform : IAA (25 : 24 : 1) and RNA were precipitated with 0.7 vol precooled ethanol and 0.2 vol 1 m acetic acid. Finally, RNA was resuspended in sterile water treated with DEPC. For Northern analysis, 300 ng total RNA per lane was fractionated on 1% agarose−1.9% formaldehyde gel and capillary blotted onto a Hybond-N membrane (Amersham).

DNA probes were labelled with [α-32P]dCTP using a random hexamer priming method (Megaprime DNA labelling system, Amersham). Phosphorimager quantifications were done with a Molecular Imager FX (Bio-Rad, Ivry Sur Seine, France).

Genomic sequencing method

The genomic sequencing method was based on that described by Clark et al. (1994), using the modifications described in Arnaud et al. (2000). The following couples of primers were used: [5S-7(-) 5′-ACATRRCDRRTRRRACCCAC-3′; 5S-8(-) 5′-RTYGGAGRGYTBTYTTTGGG-3′] or [5S-BIAIS-3 5′-ACATAACWAATAAAACCCAC-3′; 5S-BIAIS-4 5′-RTTGGAGRGTTKTTTTTGGG-3′]. R = (A or G); Y = (C or T); W = (A or T); D = (G, A or T); B = (G, T or C).

As control, pGEM-T easy vector (Promega, Charbonnières, France) containing a 5S rDNA unit was treated in the same tube as total Arabidopsis DNA, and the 5S insert was amplified from it (using primers PG1B 5′-GGGTGAATTGGGTTTGATGT-3′ and PG2B 5′-CTCCCATATAATCAACCTAC-3′) and cloned. Twenty of these clones were sequenced. The 5S rDNA unit analysed contains 34 cytosines that were all transformed to uracils, except one methylated cytosine resulting from plasmid multiplication in bacteria expressing the dcm methylation pathway (Marinus, 1987). These results demonstrated the completeness of the chemical reaction, and the methylation status along the 5S rDNA units of the different 5S arrays of the genome was authorized.

Methylation treatments

The 1037 construct was methylated by the SssI CpG Methylase or the HpaII Methylase (New England Biolabs) for 4 h at 37°C using 2 units µg−1 DNA in the recommended buffer supplemented with 80 µm S-adenosylmethionine. The reaction was repeated two and three times after the first treatment by adding fresh S-adenosylmethionine and methylase in the same proportions as the first time. Aliquots were taken after each methylation treatment, the reaction was stopped by a 15 min incubation at 65°C, and methylation efficiency was tested by digestion with HpaII endonuclease. Quantifications were performed using Molecular Analyst software (Bio-Rad).

Preparation of tobacco nuclear extract and in vitro transcription

Tobacco nuclear extracts were prepared by a modification of the procedure of Fan and Sugiura (1995) and Yukawa et al. (1997), as described by Arnaud et al. (2001).

In vitro transcription reactions from methylated 1037 constructs using tobacco nuclear extracts were done as previously described (Yukawa et al., 1997), with minor modification. Briefly, the reaction was performed in 20 µl volume containing 30 mm HEPES–KOH pH 7.9, 3 mm MgSO4, 80 mm KOAc, 0.1 mm EGTA, 2 mm DTT, 10% glycerol, 0.5 mm each of ATP, CTP, UTP, 25 µm GTP, 37 kBq [α-32P]GTP, 0.4 pmol circular plasmids, 0.5 µg ml−1α-amanitin and approximately 15 µg tobacco nuclear extract. After inoculation at 28°C for 90 min, the 32P-labelled RNA was extracted using Total RNA SafeKit (BIO101, Vista, CA, USA). The extracted RNA was separated by 5–8% polyacrylamide gel containing 7 m urea and TBE. Radioactivity was detected by Bio-Imaging Analyser BAS-2000 II (Fuji Photo Film, Tokyo, Japan).

Acknowledgements

The authors thank C. White and T. Pélissier for critical reading of the manuscript. They also thank P. Arnaud and T. Pélissier for helpful advice with the bisulfite technique, and C. Cuvillier and M.C. Espagnol for technical assistance. This work was supported by the CNRS and by the Université Blaise Pascal. O. Mathieu is the recipient of a fellowship from the Ministère de l'Enseignement supérieur et de la Recherche.

Ancillary