Plant materials:

The ddm1-1 and met1-1 mutations were generated using EMS in a Col strain background. The original mutant isolates were backcrossed at least six times to Col wild type to remove unlinked mutations (Vongs et al. 1993; Kankel et al. 2003). ddm1-1 homozygotes were examined in the first generation of self-fertilization to minimize the progressive accumulation of genetic and epigenetic alleles, as previously reported in this background (Kakutani et al. 1996). The met1-1 allele is hypomorphic but not null. The suvh4 allele corresponds to T-DNA line SALK_044606 (ABRC), the drm2 allele corresponds to T-DNA SAIL_70_E12 (ABRC), and the cmt3-11 allele corresponds to SALK_148381 (ABRC) (Chan et al. 2006). The rdr2-1 (GARLIC_1227) and dcl3-1 (SALK_005512) mutants were obtained from J. Carrington (Xie et al. 2004). The hda6-5/axe1-5 mutant was obtained from T. Guilfoyle (Murfett et al. 2001) while the nrpd2a/nrpd2b double mutant was obtained from C. S. Pikaard (Onodera et al. 2005). Natural wild-type accessions were obtained from ABRC or Lehle Seed Company; stock numbers are indicated in Table 3.

Plant material was grown on soil or 1× MS + 1% sucrose plates for 2–3 weeks before preparing for RT–PCR, cytosine methylation, or ChIP analysis. Samples from each mutant background were grown in parallel to minimize environmental effects.

RNA and DNA analysis:

DNA was isolated from rosette leaves or whole 3-week-old seedlings as previously described (Rangwala et al. 2006). RNA was isolated using TRIZOL reagent (Invitrogen, Carlsbad, CA) followed by DNAseI treatment (Invitrogen). First-strand cDNA was primed with oligo(dT)(15) primer using Superscript II reverse transcriptase (Invitrogen). Reactions excluding the reverse transcriptase were carried out in parallel to monitor DNA contamination. PCR was conducted with Taq DNA polymerase (QIAGEN, Valencia, CA) under standard cycling conditions. Primers within the transcribed region of cyclophilin (At4g38740) (Henikoff and Comai 1998) were used as PCR amplification controls. Data in Figure 1 are representative of two independent biological replicates of all samples. RT–PCR for the Sadhu elements used primers X1 + Y1 indicated in Figure 2. All primer sequences in this study are provided in supplemental Table 1 at http://www.genetics.org/supplemental/.

The 5′ ends of Sadhu3-1, Sadhu7-2, and Sadhu6-1 were determined using SMART RACE cDNA amplification (BD Biosciences) reagents and protocols and primers Sadhu3-1Y1, Sadhu7-2Y1, and Sadhu6-1X1; 3′ RACE of Sadhu3-1 used primer Sadhu3-1X3. The RACE PCR products were cloned into pGEMT-EASY (Promega, Madison, WI) and transformed into heat-shock-competent Escherichia coli using standard protocols. Plasmid from individual colonies was isolated using QIAprep spin columns (QIAGEN) and sequenced using T7 and Sp6 primers and Big-Dye Terminator cycle sequencing (Perkin-Elmer, Norwalk, CT) reagents.

DNA methylation analysis:

Genomic DNA from Col wild-type, ddm1-1, met1-1, and hda6-5 was modified by sodium bisulfite using the CpGenome DNA modification kit (Chemicon) or the EpiTect Bisulfite Kit (QIAGEN). For COBRA assays, modified DNA was first amplified by PCR (36–40 cycles) with Klentaq (CLONTECH, Palo Alto, CA) and the resulting products were digested with the endonucleases (New England Biolabs, Beverly, MA) indicated in Figure 3. Converted DNA is susceptible to PacI and HphI cleavage and resistant to TaqI, MboI, and HpaII cleavage. We controlled for efficient bisulfite conversion using a PacI COBRA assay in an amplicon of At1g01010 (primer At1g01010 F + R), which was previously determined to be unmethylated (H. Kuo, unpublished data) (Figure 3A). PCR for COBRA at the Sadhu elements used primers Bt1 + Bt2. Data in Figure 3 are representative of COBRA assays conducted on at least two independent conversions for each mutant background.

PCR products from bisulfite-modified genomic DNA templates were cloned and sequenced as described above for RACE. Primers Bt1 + Bt2 and Bb1 + Bb2 amplify “top” and “bottom” strand-specific bisulfite-modified products. Between 11 and 15 clones were sequenced per strand from Col wild type for each of Sadhu3-1, Sadhu7-2, and Sadhu6-1. Between 11 and 27 clones were sequenced per locus using either the Bt1 + Bt2 amplicon (Sadhu7-2 and Sadhu6-1) or the Bb1 + Bb2 amplicon (Sadhu3-1) generated from ddm1-1, met1-1, hda6-5, or cmt3-11 genomic templates. The sequence data were derived from at least two independent PCR amplifications per genetic background per locus. Complete bisulfite data along with DNA sequence context are presented in graphical form in supplemental Figure 2 at http://www.genetics.org/supplemental/. McrBC (New England Biolabs) digests in Table 3 were carried out at 37° overnight using the supplier's recommended conditions as described previously (Rangwala et al. 2006).

Chromatin immunoprecipitation:

Chromatin immunoprecipitation (ChIP) assays were carried out using the protocol described in Lawrence et al. (2004). Immunoprecipitation was carried out overnight using no antibody, antibody to trimethylated histone H3 lysine 4 (Abcam AB-8580, Cambridge, UK), or antibody to dimethylated histone H3 lysine 9 (Upstate 07-441, Temecula, CA). Precipitated DNA was resuspended and PCR was performed using Klentaq (CLONTECH) and standard cycling conditions for 35–40 cycles. Three independent biological replicates of chromatin immunoprecipitation were conducted. ChIP PCR used primers Sadhu3-1X1 + Y2, Sadhu7-2X2 + Y2, and Sadhu6-1X2 + Y2 for the Sadhu elements, as well as primers At4g04040 F + R (Gendrel et al. 2002) and cinful F + R as controls for H3meK4 and H3meK9 antibodies, respectively.

Derepression of Sadhu elements in different chromatin mutant backgrounds:

We recently reported a family of previously uncharacterized Arabidopsis thaliana retroelements, named Sadhu (Rangwala et al. 2006). These elements are dispersed, do not encode proteins, exist at low copy numbers, and share only moderate sequence similarity (typical pairwise nucleotide identity ∼70%). We developed an ontological scheme on the basis of sequence similarity to rename each family member (supplemental Figure 1 at http://www.genetics.org/supplemental/). One member of this family, Sadhu1-1 (At2g10410), exists in different epigenetic states in different natural accessions, despite the fact that this element in different accessions is inserted into the same chromosomal location and is nearly identical in nucleotide sequence (Rangwala et al. 2006). The silenced epigenetic state of the Sadhu1-1 allele from the Ler strain contains dense CpG methylation, and both silencing and DNA methylation can be reversed by a mutation in the DDM1 gene (Rangwala et al. 2006).

We were interested in determining whether other Sadhu elements were regulated epigenetically, as well as understanding the epigenetic mechanisms operating on the elements. One possibility is that all elements in the same transposon family are regulated using the same epigenetic mechanisms. Alternatively, different Sadhu elements might be regulated independently on the basis of their genomic environment, DNA sequence, or evolutionary history. We chose six elements that were single copy and not expressed in the Col strain (Figure 1) and examined their expression by RT–PCR in eight mutants in the Col background carrying loss-of-function or hypomorphic mutations in genes previously implicated in chromatin level gene regulation (ddm1, met1, hda6/axe1, and suvh4/kyp) or RNA-directed silencing processes (rdr2, drm2, nrpd2, and dcl3) (Figure 1). All six elements were expressed in at least one of the mutants defective in chromatin regulation. Sadhu3-1, Sadhu3-2, Sadhu5-2, and Sadhu8-1 were expressed in the ddm1, met1, and hda6 mutants. These four elements together form a group that is susceptible to multiple mutations that perturb maintenance of transcriptionally silent chromatin. In addition, Sadhu3-2 showed increased expression in the suvh4 mutant. Sadhu5-2 showed slight expression in the suvh4 mutant, while Sadhu3-1 showed slight expression in the drm2 and rdr2 mutants. By contrast, Sadhu7-2 was expressed at high levels only in the ddm1 mutant, with relatively little expression in met1, while Sadhu6-1 was expressed exclusively in the met1 mutant. Notably, none of the six Sadhu elements were strongly expressed in any of the lines carrying mutations implicated in RNA-directed de novo silencing.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.—

RT–PCR expression analysis of six Sadhu elements in Col wild-type and various chromatin or RNA silencing mutants. Cyclophilin amplification is shown as an input control.

The observation that the six Sadhu elements fell into three distinct classes with respect to chromatin mutation susceptibility (ddm1/met1/hda6, ddm1 ≫ met1, met1 only) points to the diversity of strategies used to silence these elements. We focused on representative elements from these three classes—Sadhu3-1, Sadhu6-1, and Sadhu7-2—as models to study different modes of Sadhu retroelement regulation in Arabidopsis. As a foundation for these studies, we mapped the 5′ ends of the transcripts using RACE–PCR in genetic backgrounds that expressed the element (ddm1 for Sadhu3-1 and Sadhu7-2; met1 for Sadhu6-1). In all three cases, the start of transcription mapped close to or within the sequence of the element, indicating that ectopic transcription in these mutants is not due to readthrough from a promoter far upstream of the element (Figure 2). In the case of Sadhu7-2, transcription was antisense to the direction of the retroposon and 5′ RACE analysis identified several alternate 5′ ends. The 3′ end of the Sadhu3-1 transcript was determined by 3′ RACE and shown to be polyadenylated 3′ to the poly(A) tract. Sadhu3-1 and Sadhu7-2 are embedded within DNA transposon remnants in the repeat-rich pericentromere, whereas Sadhu6-1 is found on the chromosome arm in an intergenic region between two protein-coding genes with no nearby repetitive elements (within 100 kb) (Figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.—

Detailed maps of ∼6-kb regions encompassing Sadhu3-1, Sadhu7-2, and Sadhu6-1. The direction of the centromere and the scale are indicated. The horizontal arrow below each element denotes the direction and length of the transcript, as determined by RACE–PCR. The dotted region of the arrow under Sadhu7-2 indicates a lack of a firm 5′ boundary of transcription. The 3′ end of the Sadhu6-1 transcript has not been empirically determined, as denoted by the question mark. Poly(A) or poly(T) tracts are marked on the lower strand. Target site duplications (TSD), where present, are indicated. MuDR, SINE4, REP2, and META1 are repetitive elements in the vicinity of Sadhu3-1 (Jurka et al. 2005). The location of PCR regions assayed by RT–PCR (X1 + Y1), bisulfite sequencing, and ChIP are marked for each element.

Effects of mutations in chromatin regulators on cytosine methylation at different Sadhu elements:

Next, we examined cytosine methylation at the representative Sadhu elements—Sadhu3-1, Sadhu6-1, and Sadhu7-2—both in Col wild-type and in selected mutant backgrounds. We focused our analysis on the mutations that had the most striking effect on transcription of the Sadhu elements: ddm1, met1, and hda6. We used combined bisulfite restriction analysis (COBRA) (Xiong and Laird 1997) to examine loss or gain of a restriction enzyme recognition site after genomic amplification from templates treated with sodium bisulfite, which converts unmethylated cytosines to uracil. For Sadhu3-1, which was ectopically expressed in ddm1, met1, and hda6 mutants (Figure 1), loss of TaqI cleavage indicates hypomethylation of the CpG within the TaqI restriction site. While we observed substantial TaqI cleavage in the Col wild-type sample, we noted decreased cleavage of the amplified product derived from the ddm1 mutant and little or no cleavage of the met1 mutant product (Figure 3A). However, the hda6 mutant sample showed the same amount of cleavage as Col wild type, suggesting that no change in cytosine methylation occurred in this mutant at this particular CpG site despite the loss of transcriptional silencing. We also obtained DNA sequence from within the Sadhu3-1 element from cloned products derived from Col wild-type, ddm1, met1, and hda6 bisulfite converted templates (Figure 3B; Tables 1 and 2). The ddm1 mutant samples showed a dramatic loss of CpG methylation relative to Col wild type from 89 to ∼17%, as well as a change in CpHpG methylation from 31 to 3%. The met1 mutant, consistent with the COBRA assay, showed an even greater reduction of methylation in this region, with <3.5% of cytosines methylated regardless of sequence context. The hda6 mutant also showed a dramatic decrease in CpHpG and CpHpH methylation. However, the loss of CpG methylation was more modest. Consistent with the COBRA result suggesting retention of methylation, the CpG present in the assayed TaqI site was methylated in 85% of sequenced clones. Therefore, although all three mutations—ddm1, met1, and hda6—were able to derepress the Sadhu3-1 element (Figure 1), each mutation had a unique effect on cytosine methylation at this locus.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.—

DNA cytosine methylation of Sadhu elements in Col wild-type, ddm1, met1, and hda6 mutants. (A) COBRA analysis of an unmethylated control locus, At1g01010 (cleavage confirms efficient conversion of bisulfite-modified templates), and Sadhu3-1, Sadhu7-2, and Sadhu6-1 elements. Restriction maps of assayed regions are shown to the right of the gel images. Arrowheads indicate positions of uncleaved PCR product. Numbers below lanes indicate percentage of methylation at the cytosines assayed by COBRA, as determined by sequencing of independent bisulfite-modified amplicons. (B) Percentage of methylated sites in Col wild-type, ddm1, met1, and hda6 (Sadhu3-1) genetic backgrounds by sequence context, as determined by sequencing of bisulfite-modified amplicons. Numbers in parentheses below graphs indicate the number of cytosines in each sequence context for each locus. H is C, T, or A. A complete graphical representation of bisulfite information at each individual cytosine residue is available in supplemental Figure 2 at http://www.genetics.org/supplemental/.

We next examined cytosine methylation of the Sadhu7-2 element using a COBRA assay reporting on CpG methylation at an MboI site. This site was cleaved in a majority of the products derived from Col wild-type and the hda6 mutant templates, but hypomethylated in the ddm1 and met1 mutants (Figure 3A), consistent with the accumulation of transcripts in these mutants. Sequencing of amplification products from bisulfite-treated templates revealed that CpG methylation was reduced from 95% in Col wild type to 27% in the met1 mutant and 12% in the ddm1 mutant (Figure 3B; Tables 1 and 2). CpHpG methylation was not decreased in the met1 mutant, but was greatly reduced from 57% in Col wild type to 16% in the ddm1 mutant. This more extreme hypomethylation in the ddm1 mutant correlates well with the RT–PCR data that indicated striking expression of Sadhu7-2 in the ddm1 mutant but little to no expression in the met1 or hda6 mutants (Figure 1).

The Sadhu6-1 element was also assayed for changes in cytosine methylation in the mutant backgrounds by both COBRA and bisulfite-mediated genomic sequencing. In this case, the COBRA assays monitored loss of CpG methylation at an HpaII site, as well as the creation of an HphI site upon bisulfite conversion of unmethylated DNA. Our COBRA results indicated that cytosine methylation was present in Col wild-type plants, persisted in the ddm1 and hda6 mutants, but was lost in the met1 mutant (Figure 3A). These results were corroborated by sequencing of PCR products amplified from bisulfite-treated templates (Figure 3B; Tables 1 and 2). Products derived from the ddm1 mutant showed a slight decrease in CpG methylation (99 → 83%) and a modest increase in CpHpH methylation (10 →16%) relative to Col wild type. Consistent with the COBRA result, the met1 mutant showed a stronger decrease in methylation in both the CpG (99 → 62%) and CpHpG (83 → 49%) sequence contexts relative to Col wild type. A significant reduction of methylation in the met1 but not the ddm1 mutant mirrors the expression data showing expression of the Sadhu6-1 element solely in the met1 mutant (Figure 1).

Although all three elements showed a high degree of CpG methylation (∼90%) and low levels of methylation at asymmetric CpHpH sites (∼10%), each element showed a different degree of methylation at CpHpG sites in Col wild type (Table 1). CpHpG methylation was moderate at Sadhu3-1 (24%), greater at Sadhu7-2 (45%), and highest at Sadhu6-1 (79%). Because the CMT3 chromomethylase enzyme is responsible for methylation at CpHpG sites, we examined a Col strain carrying a previously characterized hypomorphic T-DNA insertion allele in the CMT3 gene (Chan et al. 2006). We did not see any increase in expression of the Sadhu3-1, Sadhu7-2, or Sadhu6-1 loci in this Col cmt3 mutant (Figure 4A). Bisulfite sequencing of the Sadhu6-1 element in the cmt3 mutant indicated a nearly 50% reduction of CpHpG methylation, with no change in CpG methylation (Figure 4B). Therefore, the partial loss of CpHpG methylation in this particular cmt3 mutant was not sufficient to reverse silencing at this locus.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.—

Sadhu elements are not expressed in a mutant that decreases CpHpG methylation. (A) RT–PCR analysis of Sadhu3-1, Sadhu7-2, and Sadhu6-1 in a cmt3-11 mutant compared to heterozygous siblings (+/cmt3). ddm1 and met1 samples are included as controls. Cyclophilin is shown as an amplification control. (B) Percentage of methylated sites in Col wild type vs. cmt3-11, by sequence context, as determined by bisulfite-mediated genomic sequencing of Sadhu6-1. Ratios above bars are the number of methylated sites/total number of sites assayed.

Distribution of H3meK4 and H3dimeK9 at different Sadhu elements correlates with expression state:

Next we examined histone modification in nucleosomes associated with the three representative Sadhu elements using ChIP. Histone H3 trimethylated at lysine 4 (H3meK4) has been shown to be correlated with active genes in Arabidopsis and other eukaryotes, while H3 dimethylated at lysine 9 (H3meK9) is associated with silent loci (Gendrel et al. 2002; Lippman et al. 2003). The three Sadhu elements showed differential histone modifications in different mutant backgrounds. We examined a constitutively active gene, At4g04040, as a control for the H3meK4 antibody (Figure 5A), and a cinful retroelement as a control for a heterochromatic sequence (Gendrel et al. 2002) (Figure 5E). Sadhu3-1, Sadhu6-1, and Sadhu7-2 and the cinful element were associated primarily with H3meK9 in Col wild-type plants, consistent with their epigenetically suppressed states (Figure 5, B–E).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.—

Chromatin immunoprecipitation assay of histone H3 methylation at lysine 4 and lysine 9. (A) At4g04040 represents a constitutively expressed locus; (B) Sadhu3-1; (C) Sadhu7-2; (D) Sadhu6-1; and (E) cinful retroelement. I, 10% input; M, mock (no antibody); meK4, H3 trimethyl K4 antibody; meK9, H3 dimethyl K9 antibody.

The ddm1, met1, and hda6 mutations caused a shift in association from H3meK9 to H3meK4 at Sadhu3-1 (Figure 5B), accompanying the ectopic expression of this locus (Figure 1). Sadhu7-2 was expressed at high levels (Figure 1) and showed increased levels of H3meK4 relative to H3meK9 in the ddm1 mutant (Figure 5C). The strong effect of a mutation in DDM1 on histone modification state also correlated well with the cytosine methylation data showing the greatest hypomethylation at this locus in the ddm1 background (Figure 3). In the met1 mutant, which showed only slight expression of Sadhu7-2, H3meK9 remained associated with this locus. Sadhu6-1 showed a marked shift in association of H3meK4 relative to H3meK9 only in the met1 mutant (Figure 5D), consistent with expression of this locus in a met1 background (Figure 1). In the nonexpressing mutants (ddm1, hda6) H3meK9 was still associated with Sadhu6-1. Therefore, the silenced state at all three elements correlated well with the presence of H3meK9, while expression was associated with H3meK4, regardless of the genomic location, genetic background, or DNA methylation state.

Variation in expression and methylation of Sadhu elements in natural wild-type accessions:

We examined Sadhu3-1, Sadhu7-2, and Sadhu6-1 in 20 natural A. thaliana accessions for the presence of the element (using internal PCR primers), cytosine methylation (assayed by McrBC PCR), and expression (assayed by RT–PCR) (Table 3). We found a large degree of variation in all three parameters at each of the three elements. For both Sadhu6-1 and Sadhu3-1, several accessions expressed the element and silencing was not always perfectly correlated with cytosine methylation. In contrast, the Sadhu7-2 locus was not expressed in any of the wild-type accessions and was methylated in every accession in which it was present. These results suggest that Sadhu7-2, which is derepressed only in a ddm1 mutant, is more stably silenced in natural populations than the other Sadhu elements.