An Internal Rearrangement in an Arabidopsis Inverted Repeat Locus Impairs DNA Methylation Triggered by the Locus

Corresponding author: Judith Bender, Johns Hopkins University Bloomberg School of Public Health, 615 N. Wolfe St., Baltimore, MD 21205., jbender{at}mail.jhmi.edu (E-mail)

Communicating editor: B. BARTEL

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

In plants, transcribed inverted repeats trigger RNA interference (RNAi) and DNA methylation of identical sequences. RNAi is caused by processing of the double-stranded RNA (dsRNA) transcript into small RNAs that promote degradation of complementary RNA sequences. However, the signals for DNA methylation remain to be fully elucidated. The Arabidopsis tryptophan biosynthetic PAI genes provide an endogenous inverted repeat that triggers DNA methylation of PAI-identical sequences. In the Wassilewskija strain, two PAI genes are arranged as a tail-to-tail inverted repeat and transcribed from an unmethylated upstream promoter. This locus directs its own methylation, as well as methylation of two unlinked singlet PAI genes. Previously, we showed that the locus is likely to make an RNA signal for methylation because suppressed transcription of the inverted repeat leads to reduced PAI methylation. Here we characterize a central rearrangement in the inverted repeat that also confers reduced PAI methylation. The rearrangement creates a premature polyadenylation signal and suppresses readthrough transcription into palindromic PAI sequences. Thus, a likely explanation for the methylation defect of the mutant locus is a failure to produce readthrough dsRNA methylation triggers.

CYTOSINE methylation plays a critical role in directing patterns of heterochromatin formation in the genomes of mammals and plants, with effects on both gene expression and genome stability. In mammals, methylation is required for essential developmental programs including X chromosome inactivation in females and genomic imprinting (reviewed in BIRD 2002 ). In plants, methylation is required for inactivation of invasive parasitic sequences such as transposable elements (MIURA et al. 2001 ; SINGER et al. 2001 ; KATO et al. 2003 ).

One mechanism of guiding cytosine methylation to appropriate genomic loci involves an RNA signal with sequence identity to the DNA target. This process of RNA-directed DNA methylation has been well documented in plants (reviewed in MATZKE et al. 2001 ), and a potentially related process of RNA-directed heterochromatin formation occurs in the fungus Schizosaccharomyces pombe (VOLPE et al. 2002 ; SCHRAMKE and ALLSHIRE 2003 ). RNA-directed DNA methylation is interrelated with another silencing mechanism, RNA interference (RNAi). During RNAi, accumulation of double-stranded RNA (dsRNA) triggers a two-step RNA degradation process. In the first step, the dsRNA is cleaved by dicer RNAse activities to yield small interfering RNAs (siRNAs) of 25 nucleotides (nt). In the second step, these siRNAs are incorporated into an RNA-induced silencing complex and used as guides to target degradation of complementary transcripts. The observation made from plant systems that generate high levels of dsRNAs, including infecting RNA viruses and highly transcribed inverted repeat transgenes, is that RNAi triggered by the dsRNA is typically accompanied by dense methylation of DNA sequences with homology to the dsRNA precursor and its siRNA products (DALMAY et al. 2000 ; METTE et al. 2000 ; SIJEN et al. 2001 ; VAISTIJ et al. 2002 ). Furthermore, mutations in factors that control the processing of aberrant transcripts into dsRNA, such as an RNA-dependent RNA polymerase mutation, block both RNAi and DNA methylation (DALMAY et al. 2000 , DALMAY et al. 2001 ; FAGARD et al. 2000 ; MOURRAIN et al. 2000 ; BECLIN et al. 2002 ). These findings argue that dsRNA or siRNAs are likely to be the triggers for RNA-directed DNA methylation as well as RNAi. However, the mechanistic relationship between the two silencing pathways remains to be fully elucidated.

The phosphoribosylanthranilate isomerase (PAI) tryptophan biosynthetic genes in Arabidopsis provide a model system to study RNA signals for DNA methylation of relatively low-expression endogenous genes. The Wassilewskija (WS) strain of Arabidopsis carries a PAI1–PAI4 inverted repeat gene arrangement plus unlinked singlet PAI2 and PAI3 genes, and all four genes are densely methylated over their regions of sequence identity at both CG and non-CG cytosines (BENDER and FINK 1995 ; LUFF et al. 1999 ). Of the four genes, only PAI1 and PAI2 encode functional PAI enzyme, and only PAI1 is significantly expressed due to a novel unmethylated promoter that lies upstream of the PAI1 gene (MELQUIST et al. 1999 ; MELQUIST and BENDER 2003 ). The PAI2 gene is transcriptionally silenced by methylation of its more proximal promoter sequences. Mutations in WS PAI1, including missense mutations in the PAI1 coding sequence (BARTEE and BENDER 2001 ) and complete deletion of the PAI1–PAI4 inverted repeat genes (BENDER and FINK 1995 ), display a PAI-deficient blue fluorescent phenotype due to accumulation of an early intermediate in the tryptophan pathway. Transgene-induced silencing of the upstream promoter that drives PAI1 expression also confers a blue fluorescent phenotype (MELQUIST and BENDER 2003 ).

In previous work, we showed that the PAI1–PAI4 inverted repeat triggers de novo methylation of unmethylated PAI sequences (LUFF et al. 1999 ). Furthermore, we found that the inverted repeat locus is required for maintenance of dense methylation on the unlinked PAI2 and PAI3 genes; when the PAI1–PAI4 locus is deleted or segregated away through genetic crosses, the PAI2 and PAI3 genes lose most of their non-CG methylation (BENDER and FINK 1995 ; JEDDELOH et al. 1998 ; LUFF et al. 1999 ). The residual CG methylation on these loci is maintained as a relic of the previous dense methylation imprint and can be lost through subsequent rounds of DNA replication (BENDER and FINK 1995 ). The inverted repeat locus is likely to produce an RNA signal for PAI DNA methylation, because when transcription of this locus is suppressed, methylation on the PAI2 and PAI3 genes is reduced to primarily CG maintenance methylation (MELQUIST and BENDER 2003 ). However, full-length PAI transcripts are not efficiently degraded by RNAi. Furthermore, only a minority of accumulated PAI transcripts extend beyond a major polyadenylation site at the end of PAI1 to include palindromic PAI4 sequences, and although this PAI dsRNA is presumably a substrate for dicer cleavage, PAI siRNAs are not detectable by gel blot. Together, these observations lead to the view that RNA-directed DNA methylation can be promoted by much lower levels of trigger RNA species than RNAi can, but do not discriminate whether PAI dsRNA, PAI siRNA, or some other PAI-derived aberrant RNA species guides DNA methylation.

Here we describe a novel rearrangement mutation in the PAI1–PAI4 locus that creates a new premature polyadenylation site, suppresses readthrough transcripts from PAI1 into PAI4, and impairs both maintenance and de novo methylation of PAI sequences. This rearrangement mutant thus provides evidence supporting dsRNA or a processed product of dsRNA as the PAI methylation trigger.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Isolation and cloning of the WS invpai1-pai4 mutant:
Ethyl methanesulfonate (EMS)-mutagenized WS M2 seed pools were purchased from Lehle Seed, although as noted in the DISCUSSION, the nature of the rearrangement mutation recovered is inconsistent with a direct effect of EMS mutagenesis. Seeds were surface sterilized, plated on plant nutrient plus 0.5% sucrose medium (HAUGHN and SOMERVILLE 1986 ) with 0.75% agar, and screened at 2 weeks postgermination with a hand-held short wave ultraviolet (UV) light source for blue fluorescent mutants. Complementation crosses of the invpai1-pai4 mutant isolate with a blue fluorescent pai1-pai4 deletion mutant (BENDER and FINK 1995 ) indicated that the phenotype was caused by a pai1 lesion. The mutant invpai1-pai4 locus was cloned by making a DASH (Stratagene, La Jolla, CA) library from BamHI-cleaved genomic DNA and screening plaques by hybridization with a probe to direct repeat sequences that flank the inverted repeat PAI locus. The entire locus was recovered on a 17-kb BamHI fragment. Restriction mapping and sequencing of the central 6.3 kb of this fragment (GenBank no. AY357734) indicated that the mutant differed from parental WS only in the central region between PAI1 and PAI4, as described in detail in RESULTS.

DNA and RNA analysis:
A PAI1 (At1g07780) cDNA internal 0.7-kb PstI fragment probe that hybridizes to all four WS PAI genes was used for DNA and RNA gel-blot analysis of PAI sequences (BENDER and FINK 1995 ). This same sequence was used as a template for generating antisense or sense-strand PAI RNA probes. A ß-tubulin (At5g44340) cDNA probe was used as a gel loading control in RNA blot analysis. A 180-bp centromere repeat probe from plasmid pARR20-1 (gift of E. Richards, Washington University, St. Louis) was used for analysis of centromere methylation status. DNA probes were labeled using the Amersham (Buckinghamshire, UK) MegaPrime kit, and RNA probes were labeled using the Promega (Madison, WI) in vitro transcription kit and T3 polymerase. Bisulfite sequencing of methylation patterns on the upstream sequences of PAI1 and PAI2 was performed as previously described (MELQUIST and BENDER 2003 ). Polyadenylated and nonpolyadenylated RNAs were fractionated from total RNA using oligo(dT) magnetic beads (Dynal, Great Neck, NY). Ambion Millenium RNA markers were used to determine transcript length. Rapid amplification of cDNA ends (RACE) analysis was performed as previously described (MELQUIST and BENDER 2003 ) using the SMART kit (CLONTECH, Palo Alto, CA), except that a primer in the third PAI exon, P137, 5'-CACCAACAGGTTTGGCCCCACCTTCCC-3', was used for 5' RACE analysis. This primer is completely identical to all four WS PAI genes and lies outside the deleted region in WS invpai1-pai4. For reverse transcriptase-PCR (RT-PCR) analysis of PAI1 transcript 5' ends, the primers used were S15a-RTF1, 5'-GAGTACCTTGCCTCTCGAGCTCCC-3' in the first upstream exon and P129, 5'-CATCATCCTTAGGAGCTACATTC-3' in the third PAI exon.

Genetic analysis of WS invpai1-pai4:
The presence of the invpai1-pai4 rearrangement was detected in segregating populations using PCR primers that flank the pai4 deletion: PIF, 5'-CCGCCGCGTCTCTGCTGACCC-3' and PIR, 5'-GATTGGAAACAATAGGTTGATGC-3'. The primers yield a 1642-bp product from PAI2 and PAI3, a 1633-bp product from invpai1, and a 1123-bp product from pai4. Plants homozygous for the rearrangement mutation were further identified by their fluorescent phenotype. In crosses between WS and Columbia (Col), segregating PAI loci were scored using polymorphisms linked to each locus (BENDER and FINK 1995 ).

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

An internal inversion/deletion mutation in the PAI–PAI4 inverted repeat locus confers PAI-deficient phenotypes:
In the course of screening for blue fluorescent mutants in the WS strain, we isolated an unusual mutant with an internal rearrangement in the PAI1–PAI4 inverted repeat locus. The fluorescent mutant was initially identified as having a PAI1 defect by its failure to complement the fluorescent phenotype of the WS pai1-pai4 strain. Southern blot analysis of mutant genomic DNA showed that there was a partial deletion in the PAI1–PAI4 genes (see below). To understand the nature of the rearrangement in detail, we cloned and sequenced the PAI1–PAI4 locus from the mutant. This analysis revealed that the rearrangement consisted of an inversion of the central sequences in the locus together with a 519-bp deletion extending from the noncoding sequences between the two PAI genes into the middle of the PAI4 fourth exon (Fig 1). Although the overall structure of the PAI1 coding region was intact in this rearrangement, the central inversion introduced a fifth exon 9-bp deletion normally found in the PAI4 gene into the PAI1 gene. This small deletion, which removes three amino acids from the coding sequence, was previously shown to abrogate PAI enzyme function (MELQUIST et al. 1999 ). Therefore, the inversion event introduced a loss-of-function mutation into the PAI1 coding sequence, accounting for the fluorescent PAI-deficient phenotype of the mutant.

View larger version (17K): In this window In a new window Download PPT slide   Figure 1. A model for the genesis of the invpai1-pai4 rearrangement. The solid vertical arrow represents PAI1 and the shaded vertical arrow represents PAI4, with the end of the PAI1 sequences in the central region indicated by a horizontal line. Small arrows indicate points of breakage and rejoining for (A) an inversion between the 3' ends of PAI1 and PAI4 and (B) a deletion of part of the inverted region. The deleted region is indicated by a dashed line. On the basis of the positions of PAI1 vs. PAI4 polymorphisms in invpai1, the inversion breakpoint occurred somewhere in the 421-bp region between the 3' end of the third exon and the 5' end of the fifth exon, consistent with the same breakpoint defining the end of the deleted sequences in the pai4 fourth exon.

The inversion event also changed the sequences immediately downstream of PAI1. In WS, the PAI1–PAI4 inverted repeat is centrally asymmetric, with the 3' untranslated sequences downstream of PAI1 extending for 263 bp before colliding with the 3' untranslated sequences downstream of PAI4, which extend for only 20 bp (MELQUIST et al. 1999 ). This central asymmetry was inverted in the rearrangement mutant so that PAI1 now had only 20 bp of PAI 3' untranslated sequences before colliding with PAI4 3' untranslated sequences, which include 246 bp of this region before the deletion junction. Overall, the rearrangement altered the palindromic structure of the locus so that the palindromic arms were shortened by 500 bp and the central nonpalindromic sequences were increased by 500 bp relative to the WS structure. We subsequently refer to the inversion/deletion rearrangement allele as WS invpai1-pai4.

The WS invpai1-pai4 rearrangement mutant displays reduced PAI DNA methylation and silencing:
WS invpai1-pai4 mutant genomic DNA was tested for PAI methylation changes by both Southern blot and genomic sequencing assays. Southern blot analysis with the methylation-sensitive isoschizomers HpaII and MspI revealed that the mutant DNA had partially reduced methylation at all three PAI loci (Fig 2). HpaII and MspI both cleave the sequence 5'-CCGG-3', but HpaII is inhibited by methylation of either the inner (CG) or the outer (CCG) cytosine whereas MspI is inhibited only by methylation of the outer cytosine. Each PAI locus contains a single HpaII/MspI site in the second intron, with flanking sites in unmethylated sequences at different distances from the central site for each locus (BENDER and FINK 1995 ; LUFF et al. 1999 ). In WS, the three PAI loci are highly refractory to cleavage by either HpaII or MspI, diagnostic of dense CG and CCG methylation of the recognition site. In the WS invpai1-pai4 mutant, all three loci displayed increased cleavage by HpaII and MspI. In contrast, there was no difference between wild type and mutant in HpaII/MspI cleavage patterns at methylated centromere repeat sequences, indicating that the methylation changes are specific to PAI sequences. The previously characterized WS pai1-pai4 mutant (BENDER and FINK 1995 ), with a complete deletion of the inverted repeat PAI genes, displayed a similar pattern of increased HpaII and MspI cleavage specifically at PAI loci.

View larger version (32K): In this window In a new window Download PPT slide   Figure 2. The invpai1-pai4 rearrangement confers reduced density of PAI DNA methylation. Genomic DNA samples prepared from WS, WS invpai1-pai4 (inv), or WS pai1-pai4 () were digested with the methylation-sensitive isoschizomers HpaII (H) and MspI (M). (Top) Samples were probed with a PAI internal cDNA fragment. Methylated bands are denoted with asterisks. P1–P4 is PAI1–PAI4, inv is invpai1-pai4, P2 is PAI2, and P3 is PAI3. (Bottom) The same samples were probed with a 180-bp centromere repeat probe (CEN).

To understand PAI methylation patterning in more detail, we performed sodium bisulfite genomic sequencing on mutant DNA in the regions upstream of PAI1 or PAI2, extending from flanking heterologous sequences unique to each gene into PAI-identical proximal promoter sequences. Previous sequencing of the same regions in WS showed that the PAI-identical regions of both genes are densely methylated at CG and non-CG cytosines with very little spread into the flanking heterologous sequences (LUFF et al. 1999 ). In the WS invpai1-pai4 mutant, we found that methylation in the PAI-identical region was reduced for both PAI1 and PAI2, with a strong loss of non-CG methylation (Fig 3). This pattern is similar to that previously observed on the PAI2 gene when the PAI1–PAI4 locus was replaced with a singlet PAI1 gene crossed in from the PAI-unmethylated Col strain background (Hyb4 in LUFF et al. 1999 ), when the PAI1–PAI4 locus is deleted (JEDDELOH et al. 1998 ) or when transcription through the PAI1–PAI4 inverted repeat is suppressed by targeted methylation of the upstream promoter region (MELQUIST and BENDER 2003 ). Because non-CG methylation is diagnostic of RNA-directed DNA methylation (PELISSIER et al. 1999 ), the loss of non-CG methylation in the invpai1-pai4 mutant implies a defect in an RNA signal.

View larger version (22K): In this window In a new window Download PPT slide   Figure 3. Genomic bisulfite methylation sequencing data for PAI1 and PAI2 proximal promoter regions in invpai1-pai4. Eight independent top-strand clones were sequenced for PAI1 or PAI2. The percentage of 5-methyl-cytosines out of total cytosines sequenced within the region of PAI sequence identity (344 bp for PAI1 or 338 bp for PAI2) is shown, divided into the contexts CG (solid), CNG (open), and other contexts (shaded). For comparison, previously determined wild-type WS PAI1 and PAI2 data are shown (LUFF et al. 1999 ).

The WS invpai1-pai4 mutant, with a defective pai1 gene product, is relatively weakly fluorescent with only modest effects on plant morphology and fertility (Fig 4). In contrast, a WS pai1 missense mutant, with a defective pai1 gene product and a heavily methylated and silenced PAI2 gene, is strongly fluorescent in all parts of the plant and has reduced size and fertility relative to the parental WS strain (BARTEE and BENDER 2001 ). This comparison suggests that the reduced methylation on PAI2 in the invpai1-pai4 mutant partially relieves its transcriptional silencing and partially compensates for the pai1 defect, as we previously showed in other strain backgrounds in which PAI1 activity is compromised and PAI2 is demethylated (BENDER and FINK 1995 ; JEDDELOH et al. 1998 ; BARTEE and BENDER 2001 ; BARTEE et al. 2001 ; MALAGNAC et al. 2002 ). However, we could not directly detect increased steady-state message levels for the PAI2 gene in the WS invpai1-pai4 mutant vs. parental WS because in both strains PAI2 expression is masked by stronger expression from the PAI1 locus (see below).

View larger version (26K): In this window In a new window Download PPT slide   Figure 4. The invpai1-pai4 rearrangement confers a PAI-deficient blue fluorescent phenotype. (Left) Diagrams indicate the functional status, methylation density, and transcriptional activity of the PAI1 and PAI2 genes in invpai1-pai4 and other mutants with defects at the PAI1–PAI4 locus, with wild-type WS for comparison. PAI genes that encode functional enzyme are indicated by solid arrows, and PAI genes that encode nonfunctional enzymes are indicated by shaded arrows. Direct repeats that flank the PAI1–PAI4 locus are shown as red arrows. Boxes around the PAI genes indicate DNA methylation, with a solid line indicating dense CG and non-CG methylation and a dashed line indicating reduced, mostly CG methylation. Arrowheads indicate transcription, and X's indicate transcriptional silencing. (Right) Visible (left) and UV (right) light photographs of representative 2-week-old seedlings of the indicated genotypes.

The weak PAI-deficient phenotypes of the WS invpai1-pai4 mutant were similar to those observed for the WS pai1-pai4 mutant, which has a complete deletion of the inverted repeat and reduced PAI2 methylation (Fig 4; BENDER and FINK 1995 ). However, the invpai1-pai4 mutant phenotypes were stable, with no nonfluorescent revertants detected out of thousands of plants screened, whereas the pai1-pai4 mutant phenotypes were only semistable, with 1–5% nonfluorescent and PAI-demethylated progeny resulting in each generation of self-pollination (BENDER and FINK 1995 ). This difference suggests that the partial deletion of PAI1–PAI4 is able to maintain residual methylation of the PAI2 gene better than the complete deletion of the locus. The weak PAI-deficient phenotypes of the WS invpai1-pai4 mutant were also similar to those observed for WS carrying a transgene S15aIR that triggers methylation and silencing of the upstream promoter that drives PAI1 transcription, with a concomitant reduction in PAI2 methylation (Fig 4; MELQUIST and BENDER 2003 ). Like the WS invpai1-pai4 mutant, the WS(S15aIR) strain is stably fluorescent, implying that this strain can maintain residual PAI2 methylation, in contrast to a complete deletion of PAI1–PAI4.

The WS invpai1-pai4 rearrangement does not affect transcription initiation or 5' processing of the PAI1 transcript:
To understand the effects of the WS invpai1-pai4 rearrangement mutation on PAI transcripts, we performed both RNA gel-blot analysis and RACE PCR analysis of PAI RNAs. Previous analysis of PAI transcripts in parental WS showed that PAI1 is the only significantly expressed gene (MELQUIST et al. 1999 ; MELQUIST and BENDER 2003 ). Accumulated PAI1 transcripts include a major correctly spliced and polyadenylated transcript of 1200 nt, a polyadenylated species of 1900 nt, and low levels of longer polyadenylated transcripts (MELQUIST and BENDER 2003 ; Fig 5). RNA gel-blot analysis with a PAI cDNA probe showed that the WS invpai1-pai4 mutant accumulated a major broad band of transcripts ranging from 1000 to 1200 nt (Fig 5A). This major transcript population was 2.5-fold more abundant than the major 1200 nt transcript population in WS. In addition, the mutant accumulated an 1900-nt species. As in WS, the mutant PAI RNAs were recovered predominantly in the polyadenylated fraction.

View larger version (62K): In this window In a new window Download PPT slide   Figure 5. PAI steady-state transcript analysis in the invpai1-pai4 mutant. (A) Total RNA (total) from WS or invpai1-pai4 (inv) was separated into polyadenylated [poly(A)+] and nonpolyadenylated [poly(A)-] fractions, gel blotted, and probed with a PAI probe. The blot was stripped and reprobed with ß-tubulin (TUB) to control for loading of poly(A)+ species. The ethidium-bromide (EtBr)-stained gel is shown to control for loading of poly(A)- species. Molecular weight markers in kilonucleotides are indicated in the left margin. (B) Polyadenylated RNA prepared from the indicated strains was used as a template for RT-PCR amplification of PAI1 transcript 5' ends. An ethidium-bromide-stained gel of each RT-PCR reaction is shown, with molecular weight (MW) markers in leftmost lane, sizes of markers in base pairs in the left margin, and the identities of sequenced products indicated in the right margin. In the diagrams of PAI splice variants, the first upstream exon is represented by a shaded box, the second 26-nt upstream exon is represented by a solid box, and PAI exons are represented by open boxes. Lines indicate intron sequences that are retained in the longest splice variant. Note that in Cvi-0, the first upstream intron is 26 nt shorter than in WS due to two small deletions. (C) Duplicate blots of polyadenylated RNA prepared from the indicated strains electrophoresed in parallel on the same gel were hybridized with either a PAI antisense-strand (AS) or a PAI sense (S)-strand RNA probe. Both a short (AS-S) and a long (AS-L) exposure of the same antisense-strand-probed blot are shown, in the first case to resolve the predominant lower molecular weight species and in the second case to clearly show the less-abundant higher molecular weight species. A single exposure of the sense-strand-probed blot is shown. A nonspecific band detected by the sense-strand probe is marked with an asterisk in the right margin. This band serves as an internal loading control for the sense-strand blot. In addition, reprobing of both blots with a TUB probe indicated approximately equal loading of all samples (data not shown).

For RACE analysis, the 5' or 3' ends of PAI cDNAs were amplified by RT-PCR from WS invpai1-pai4 RNA using an anchoring primer at the transcript end plus an internal PAI gene-specific primer. Individual PCR products were cloned and sequenced. The PAI gene-specific primers were designed to avoid regions that contain polymorphisms among the various WS PAI genes and also to avoid the deleted region in pai4, so that transcripts from any of the PAI genes could potentially be amplified. The RACE analysis showed that all the detectable transcripts in the WS invpai1-pai4 mutant corresponded to PAI1 (Table 1; Table 2). This result indicates that even if other PAI genes such as PAI2 are reactivated by the reduced methylation in the mutant, PAI1 transcripts still accumulate to the highest steady-state levels.

View this table: In this window In a new window   Table 1. 5' RACE determination of PAI transcript ends for WS invpai1-pai4

View this table: In this window In a new window   Table 2. 3' RACE determination of PAI transcript ends for WS invpai1-pai4

5' RACE analysis showed that mutant PAI1 transcripts initiated in the same region and had the same two types of splice variant structures as previously determined by 5' RACE for parental WS (MELQUIST and BENDER 2003 ). As in WS, WS invpai1-pai4 transcripts initiated at an upstream promoter derived from duplicated sequences of the S15a ribosomal protein gene that lies 500 bp upstream of the PAI methylation boundary and 850 bp upstream of the PAI1 translational start codon (Table 1). The majority class of transcripts had 706 nt of upstream sequences removed by a single splicing event between the S15a first intron donor site and a cryptic acceptor site 96 nt upstream of the PAI1 translational start codon. The other 5' variant was spliced twice in the upstream region to retain a central 26-nt exon. In this variant, the S15a first intron donor site was joined to the S15a first intron acceptor site to remove 416 nt of intron sequences, and then a cryptic donor site was joined to the cryptic acceptor site at -96 nt to remove 264 nt of intron sequences.

Given that the processing of the PAI1 upstream sequences involves cryptic splicing events and that the size of a transcript that failed to splice the upstream sequences would correspond to the 1900-nt species detected by RNA gel-blot analysis in WS invpai1-pai4 and WS, we tested explicitly for the presence of this long 5' splice variant using RT-PCR designed to optimize detection of the long lower-abundance species (Fig 5B). Specifically, we used a forward primer in the S15a first exon with a reverse primer in the PAI third exon, such that the putative long 5' splice variant would yield a relatively short PCR product. These primers are predicted to yield 308- and 334-bp products from the two upstream-spliced variants, a 1014-bp product from a species in which the upstream sequences are unspliced but the PAI first and second introns are correctly spliced, and a 1423-bp product from an unspliced template. As a control, we tested RNA prepared from the Col strain, which has a single unmethylated PAI1 gene and lacks a detectable 1900-nt PAI species in RNA gel-blot analysis. We also tested RNA prepared from the Cape Verde Islands (Cvi-0) strain, which carries a methylated PAI inverted repeat in which the PAI1 and PAI4 genes lie 839 bp farther apart than in WS (MELQUIST et al. 1999 ). Like WS invpai1-pai4 and WS, Cvi-0 produces a detectable 1900-nt transcript species, plus unique higher molecular weight species (Fig 5C). The RT-PCR analysis revealed that all four strains yielded short products corresponding to the upstream-spliced transcripts and a species at 600 bp; however, only the three strains with methylated PAI inverted repeats yielded detectable amounts of an 1000-bp product. Cloning and sequencing of RT-PCR products showed that the 600-bp product corresponded to a fortuitously amplified catalase 2 (At4g35090) transcript and that the 1000-bp product corresponded to a PAI1 transcript that failed to splice either the S15a first intron or the cryptic intron in upstream sequences, but that correctly spliced the PAI first and second introns (Fig 5B). Other RT-PCR products were not analyzed further. The RT-PCR analysis thus shows that the 1900-nt species detected in strains with methylated PAI inverted repeats corresponds to a PAI1 5' splice variant that fails to splice the upstream sequences between the S15a transcript start and more proximal PAI sequences. Presumably, there was a bias against recovering this long lower-abundance alternatively spliced species under the conditions we used for 5' RACE analysis. We also tried to detect the 1900-nt species using either total or polyadenylated RNA gel-blot analysis with a probe corresponding to the upstream intron sequences, but this approach did not give a signal above background hybridization (data not shown), perhaps because the intron probe sequences contain a number of poly(T) tracts that can cross-hybridize to polyadenylated transcripts.

The WS invpai1-pai4 rearrangement produces a novel prematurely polyadenylated transcript from the invpai1 gene and suppresses readthrough transcription into palindromic pai4 sequences:
3' RACE analysis of WS invpai1-pai4 RNA revealed a novel type of PAI1 transcript relative to parental WS (Table 2; MELQUIST and BENDER 2003 ). In this species, the fourth intron failed to splice and the transcript was polyadenylated internal to the fourth intron sequences. The novel species is 200 nt shorter than the full-length WS PAI1 transcript and therefore likely corresponds to the lower molecular weight shift detected by gel-blot analysis of total mutant PAI transcripts (Fig 5A). A second type of PAI1 3' end species was also detected: a full-length correctly spliced PAI transcript that was polyadenylated 70 nt downstream of the translational stop codon (Table 2).

In previous analysis, we found that the longest PAI transcripts in WS and Cvi-0 hybridize to a PAI sense-strand-specific RNA probe, indicating that they are 3' readthrough transcripts into palindromic PAI4 sequences (MELQUIST and BENDER 2003 ; Fig 5C). The shift upwards in the readthrough transcripts in Cvi-0 vs. WS reflects the different spacing between PAI1 and PAI4 in the two strains. To similarly determine whether the invpai1-pai4 mutant transcript population included readthrough transcripts with palindromic PAI4 sequences, which might be selected against during 3' RACE analysis, we performed RNA gel-blot analysis of polyadenylated RNA with a sense-strand-specific RNA probe made on a PAI cDNA template (Fig 5C). The probe (MATERIALS AND METHODS) has 347 bp of homology with the PAI4 sequences remaining in the pai4 rearrangement. As a control for nonspecific hybridization, we also analyzed RNA prepared from the Col strain, which completely lacks PAI4 inverted repeat sequences. In WS invpai1-pai4 RNA, the sense-strand probe hybridized to a nonPAI species only at 800 nt, as also observed in other samples tested, including Col. Thus, the WS invpai1-pai4 mutant does not produce detectable transcripts that contain palindromic PAI4 material.

The WS invpai1-pai4 locus can receive dense methylation triggered by the parental WS PAI1–PAI4 locus, but cannot maintain dense methylation in its absence:
In previous work we showed that the WS PAI1–PAI4 locus is a potent trigger of dense CG and non-CG PAI methylation (LUFF et al. 1999 ). To determine whether the rearranged invpai1-pai4 locus can be a target for this dense methylation, we generated invpai1-pai4/PAI1–PAI4 F₁ heterozygotes by crossing WS invpai1-pai4 to parental WS, either with the mutant as the male parent and WS as the female parent or vice versa. DNA prepared from cauline leaves of individual F₁ plants was analyzed by HpaII/MspI Southern blot relative to DNA prepared from the parental strains to assess changes in PAI methylation patterns. Regardless of whether the mutant was the male or the female parent in the cross, F₁ plants yielded DNA that showed increased resistance of the invpai1-pai4 locus to cleavage by HpaII relative to the invpai1-pai4 homozygous parent DNA (Fig 6A). Similarly, the PAI2 locus showed increased resistance to HpaII cleavage. These patterns indicate that the partially demethylated loci inherited from the mutant parent rapidly acquire increased methylation during the F₁ generation.

View larger version (39K): In this window In a new window Download PPT slide   Figure 6. The invpai1-pai4 rearrangement can receive dense methylation conferred by the intact PAI1–PAI4 locus in a heterozygote, but cannot maintain it once the PAI1–PAI4 locus is removed by segregation. Genomic DNA samples prepared from the indicated strains were digested with the methylation-sensitive isoschizomers HpaII (H) and MspI (M) and probed with a PAI internal cDNA fragment. Methylated bands are denoted by asterisks. P1–P4 is PAI1–PAI4, inv is invpai1-pai4, P2 is PAI2, and P3 is PAI3. (A) Southern blot analysis of DNA from WS, WS invpai1-pai4 (inv), or representative F1 hybrids between these two strains made with the mutant as either the male or the female parent are shown. Similar results were obtained in three independent experiments. Note that the F1 samples were run on the same blot as the parental samples, but a longer exposure of these samples is shown to adjust for loading differences. (B) Southern blot analysis of DNA from WS, WS invpai1-pai4 (inv), or F2 plants with the indicated genotypes at the PAI1–PAI4 locus segregated from a WS x invpai1-pai4 F1 parent are shown.

Interestingly, the PAI3 locus reproducibly showed a different pattern of remethylation. For F₁ plants generated with the mutant as the male parent, this locus displayed increased resistance to HpaII cleavage, similarly to the other PAI loci. However, for F₁ plants generated with the mutant as the female parent, PAI3 retained susceptibility to HpaII cleavage similar to that observed in the invpai1-pai4 mutant parent. This pattern suggests that the wild-type PAI1–PAI4 locus is not an efficient trigger of PAI3 methylation when inherited from the male parent in a cross and ties in with two previous observations. First, in experiments in which demethylated PAI genes are remethylated, PAI3 is typically slower to remethylate than the other loci (LUFF et al. 1999 ; MALAGNAC et al. 2002 ). Second, in experiments where the WS PAI1–PAI4 locus is used to trigger de novo methylation of the allelic unmethylated Col PAI1 gene in F₁ heterozygous plants, Col PAI1 is not efficiently methylated when the WS PAI1–PAI4 locus is inherited from the male parent in the cross (LUFF et al. 1999 ). A possible explanation for the different potency of the PAI1–PAI4 methylation trigger when inherited from the male parent vs. the female parent is suppressed transcription of the paternally inherited genome during the early cell divisions in the F₁ embryo (VIELLE-CALZADA et al. 2000 ); in this view, there would be a lag in the production of methylation-triggering RNA from the PAI1–PAI4 locus when inherited from the male.

To further determine whether dense methylation can be maintained on the mutant invpai1-pai4 locus after the PAI1–PAI4 locus is segregated away, DNA prepared from pooled homozygous invpai1-pai4/invpai1-pai4 F₂ progeny segregated from F₁ plants made with the mutant as the male parent was tested by HpaII/MspI Southern blot assay. This analysis showed that the segregated F₂ homozygotes reverted to the partially methylated pattern seen in the invpai1-pai4 parent (Fig 6B). In contrast, DNA prepared from pooled invpai1-pai4/PAI1–PAI4 heterozygous F₂ sibling plants maintained dense methylation on all three PAI loci at a slightly higher level than that observed in the F₁ parent, as indicated by increased resistance to MspI cleavage. This pattern is consistent with a progressive increase in methylation upon inbreeding in the presence of the intact PAI1–PAI4 locus. Similar results were obtained for methylation analysis of individual F₂ plants of each genotype (data not shown).

The WS invpai1-pai4 locus cannot trigger de novo methylation of an unmethylated PAI2 target gene:
As another test for whether the rearranged invpai1-pai4 locus could produce a PAI methylation signal, we asked whether the locus could trigger de novo methylation of an unmethylated PAI2 gene. When the parental WS PAI1–PAI4 locus is combined via genetic crosses with an unmethylated PAI2 gene from the Col strain, the Col PAI2 gene becomes methylated de novo, achieving a methylation density similar to that seen for WS PAI2 by the F₃ or F₄ generation of inbreeding by self-pollination (LUFF et al. 1999 ). Similarly, when a WS pai1 missense mutant pai1-PAI4 locus is combined with an unmethylated PAI2 gene from the Landsberg erecta (Ler) strain, the Ler PAI2 gene becomes methylated de novo, with a progressive increase in methylation upon inbreeding (MALAGNAC et al. 2002 ). In this second case, because of the defect in PAI1 enzyme, the progressive accumulation of PAI2 methylation and transcriptional silencing can be monitored by visual inspection for a blue fluorescent PAI-deficient phenotype.

To test the invpai1-pai4 locus for de novo methylation of PAI2, we crossed the mutant with wild-type Col and identified seven F₂ individuals that were homozygous for the invpai1-pai4 locus and homozygous for the PAI2 gene inherited from Col. None of these plants was fluorescent when newly segregated, and representative lines did not acquire a fluorescent phenotype diagnostic of PAI2 silencing even after three generations of inbreeding by self-pollination to the F₅ generation. HpaII/MspI Southern blot analysis did not reveal any evidence of de novo PAI2 methylation (Fig 7). A control cross between the WS pai1 missense mutant and wild-type Col generated eight F₂ individuals that were homozygous for the pai1-PAI4 mutant locus and homozygous for the PAI2 gene inherited from Col. All eight plants were fluorescent when newly segregated. By the F₅ generation, representative inbred lines were phenotypcially similar to the WS pai1 missense mutant (Fig 4) and displayed acquisition of de novo PAI2 methylation as monitored by HpaII/MspI Southern blot analysis (Fig 7). These results indicate that the invpai1-pai4 locus cannot trigger new methylation on a PAI target sequence, in contrast to the unrearranged PAI1–PAI4 or pai1-PAI4 loci.

View larger version (70K): In this window In a new window Download PPT slide   Figure 7. The invpai1-pai4 rearrangement is defective for de novo methylation of a PAI2 methylation target inherited from Col. Genomic DNA samples prepared from WS pai1 (pai1), WS invpai1-pai4 (inv), F5 generation tissue from a representative invpai1-pai4/invpai1-pai4 Col PAI2/Col PAI2 hybrid (invxCol), F5 generation tissue from a representative WS pai1-PAI4/WS pai1-PAI4 Col PAI2/Col PAI2 hybrid (pai1xCol), and wild-type Col were digested with the methylation-sensitive isoschizomers HpaII (H) and MspI (M) and probed with a PAI internal cDNA fragment. Methylated bands are denoted by asterisks. P1–P4 is PAI1–PAI4, inv is invpai1-pai4, P2 is PAI2, and P3 is PAI3.

It should be noted that in the F₅ inbred plants from the control cross between the WS pai1 missense mutant and wild-type Col, the Col PAI2 locus did not achieve fully dense methylation relative to the parental WS PAI2 locus (Fig 7). A similar pattern was previously observed for the analogous experiment done with a cross between the WS pai1 missense mutant and wild-type Ler (MALAGNAC et al. 2002 ). This resistance of PAI2 to full methylation could reflect a selection bias against the most strongly silenced PAI-deficient cells in the pai1 mutant background.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Cytosine methylation is an important regulator of gene expression and genome stability in mammals and plants. A key question is how methylation is targeted to specific regions of the genome. In plants, sequences that produce dsRNA, including RNA viruses and transcribed inverted repeats, can be potent triggers for methylation of homologous genomic DNA sequences. For example, the transcribed endogenous PAI1–PAI4 inverted repeat in the WS strain of Arabidopsis triggers dense methylation on PAI target sequences at unlinked positions in the genome (LUFF et al. 1999 ). The ability of this inverted repeat to signal methylation is suppressed by silencing the upstream promoter that drives transcription through the locus, indicating an RNA-based DNA methylation mechanism (MELQUIST and BENDER 2003 ). In this work, we characterized invpai1-pai4, a mutant variant of WS with an intact promoter region but an internal deletion and rearrangement of the central sequences in the PAI1–PAI4 locus (Fig 1). The mutant locus confers reduced density of methylation on WS PAI sequences without affecting methylation patterning elsewhere in the genome (Fig 2; Fig 3). Moreover, the mutant locus is defective in triggering de novo methylation on a previously unmethylated PAI target sequence (Fig 7). The rearrangement activates a novel premature polyadenylation site and suppresses readthrough into palindromic PAI4 sequences (Table 2; Fig 5). These findings support the view that the PAI methylation defect in the rearrangement is conferred by reduced production of a dsRNA trigger for methylation due to an altered PAI polyadenylation profile.

In previous work, we found considerable natural variation in PAI1–PAI4 inverted repeat structures among WS and other wild isolates of Arabidopsis (MELQUIST et al. 1999 ). Variations included different amounts of central sequences separating PAI1 and PAI4 and flanking duplications that could be explained as gene conversion events. The invpai1-pai4 rearrangement characterized here likely reflects the genetic instability of the locus. This rearrangement was isolated from an EMS-mutagenized seed population, but presumably occurred either as an indirect consequence of the mutagenesis or as an unrelated event. A possible explanation for the mutant structure is that an initial inversion event occurred by pairing and homologous recombination between PAI1 and PAI4, followed by a deletion event extending from the recombination breakpoint in PAI4 into the 3' end of the gene (Fig 1). Interestingly, the deletion breakpoint at the 3' end of the gene is 3 bp from the point where the palindromic and potentially aligned sequences between PAI1 and PAI4 end. It should be noted that an inversion between PAI1 and PAI4 without the accompanying central deletion and consequent reduced PAI2 methylation and transcriptional silencing would likely yield a severe PAI-deficient mutant by introducing the inactivating fifth exon mutation into the expressed PAI1 gene. In fact, such a mutant could potentially be an embryo-lethal tryptophan auxotroph. Thus, inversions at this locus might occur at a relatively high frequency, but not be recovered as viable plants.

The PAI1 promoter region presents a complex arrangement of a proximal PAI promoter fused to an upstream S15a-derived promoter. In WS and WS invpai1-pai4, in which the proximal PAI promoter sequences are methylated, only transcripts that initiate at the upstream promoter are detected by 5' RACE (MELQUIST and BENDER 2003 ; Table 1). In contrast, in Col, in which the proximal PAI promoter sequences are unmethylated, both upstream and proximal transcript starts are detected by 5' RACE (MELQUIST and BENDER 2003 ). Although we have not quantified the levels of transcripts initiating at each start site in Col, two lines of evidence suggest that the more proximal site is used preferentially when the promoter region is unmethylated. First, RNA gel-blot analysis shows that Col has an overall lower accumulation of PAI transcripts than does WS (BENDER and FINK 1995 ; Fig 5C). Furthermore, on the basis of cDNA abundance and RACE analysis, approximately half of the accumulated Col PAI transcripts correspond to PAI2 species initiating at the proximal PAI promoter (MELQUIST et al. 1999 ; MELQUIST and BENDER 2003 ). These data can be accounted for by proposing that most Col PAI1 transcripts initiate at the proximal PAI promoter even though this promoter is weaker than the distal S15a promoter. Second, the low-abundance 1900-nt splice variant that initiates at the upstream start site is not detected in Col either by RNA gel blot or by RT-PCR (Fig 5), suggesting lower expression from this upstream start site when the proximal start site is unmethylated and available than in strains like WS and Cvi-0 in which the proximal start site is methylated and silenced.

The invpai1-pai4 mutant illustrates that downstream sequences can influence the choice of splicing events in a plant transcript. In this mutant, all the sequences upstream of the invpai1 fifth exon 9-bp deletion introduced by the inversion event are identical to those in the wild-type PAI1 transcript, and yet uniquely in the mutant the fourth exon fails to splice at a high frequency (Table 2). Therefore, the fifth exon polymorphism and/or the novel rearranged sequences downstream of the invpai1 stop codon contribute to the efficiency of fourth intron splicing. The retention of fourth intron sequences has the consequence of revealing a new polyadenylation region, resulting in the novel prematurely terminated transcripts. In those mutant transcripts that do splice out the fourth intron, polyadenylation occurs 70 bp downstream of the translational stop codon (Table 2) vs. 130 bp downstream of the translational stop codon in wild-type WS (MELQUIST and BENDER 2003 ).

Several interrelated mechanisms could contribute to the increased PAI transcript accumulation observed in the invpai1-pai4 mutant vs. wild-type WS (Fig 5). First, the novel polyadenylation sites used in the invpai1 transcripts (Table 2) could create messages that are more resistant than the wild-type PAI1 species to RNA degradation. Second, the population of long readthrough transcripts detected in WS (Fig 5C) is presumably converted to shorter species in the mutant by activation of new polyadenylation signals, such that the heterogeneous long population is collapsed down into a more homogenous short population. This "collapsed" population is more likely than the original dsRNA population to accumulate to higher steady-state levels because it is no longer a substrate for dicing into siRNAs. Third, low levels of siRNAs produced by dicing readthrough transcripts could contribute to a low level of PAI transcript degradation via RNAi, and this effect would be blocked by readthrough suppression. Once dicer mutants that affect siRNA production are identified in Arabidopsis (FINNEGAN et al. 2003 ), we can use such mutants to determine the contribution of RNAi to PAI steady-state message levels in WS.

When transcription of WS PAI1–PAI4 is suppressed by a transgene that triggers methylation and silencing of the S15a-derived promoter sequences upstream of PAI1, methylation on the singlet genes PAI2 and PAI3 is reduced, but methylation on the PAI–PAI4 inverted repeat itself is not significantly affected (MELQUIST and BENDER 2003 ; Fig 4). In contrast, the invpai1-pai4 mutant displays reduced methylation on all PAI loci, including the inverted repeat (Fig 2 Fig 3 Fig 4). We previously proposed two models to account for the findings in the WS(S15aIR) promoter silenced strain: either the PAI1–PAI4 inverted repeat maintains its own methylation via an RNA-independent mechanism such as a DNA structure signal or the PAI1–PAI4 inverted repeat is more sensitive than the singlet PAI genes to RNA-based methylation signals (MELQUIST and BENDER 2003 ). In the second model, the transcriptional activity of PAI1–PAI4 and/or the unusual structure of the locus could contribute to its stronger susceptibility. The findings for the invpai1-pai4 mutant can be accommodated by either model. In the first case, the alteration in the overall palindromic structure of the mutant locus could affect a methylation mechanism that responds to intrinsic structural cues. In the second case, stronger suppression of the RNA signal in the invpai1-pai4 mutant vs. the promoter-silenced strain WS(S15aIR) could drop the RNA signal below the threshold needed to maintain methylation at the inverted repeat; furthermore, the altered structure of the locus could reduce its interactions with RNA.

The PAI1–PAI4 locus provides general insights into how methylation patterns are established and maintained on endogenous methylation targets such as transposable elements. Our previous work showed that this single locus can generate an aberrant RNA signal sufficient for methylation of all related sequences in the genome, even though the aberrant RNA is not sufficient for effective RNAi (LUFF et al. 1999 ; MELQUIST and BENDER 2003 ). By analogy, for dispersed transposable elements, only one or a few transcribed elements could generate an RNA signal sufficient for methylating and silencing the entire group of related sequences. Here, our studies of the invpai1-pai4 mutant show that methylation imprints established by an RNA signal are difficult to completely remove, even when the methylation trigger locus is rearranged to severely suppress the production of the signal. In particular, in contrast to a mutant with complete deletion of the PAI1–PAI4 locus (BENDER and FINK 1995 ), the invpai1-pai4 mutant does not yield PAI-demethylated progeny at a detectable frequency. The stability of the residual methylation in invpai1-pai4 therefore implies that low levels of an RNA trigger for DNA methylation persist in this mutant. In fact, the levels of the RNA trigger are likely to be extremely low for the following reasons: PAI dsRNA cannot be detected in the mutant by RNA gel blot (Fig 5C); the increase in steady-state levels of PAI transcripts suggests a loss of PAI-directed RNAi (see above); the loss of non-CG methylation patterning (Fig 2 and Fig 3) diagnostic of RNA-directed DNA methylation (PELISSIER et al. 1999 ) suggests an RNA signal deficiency; and the mutant locus is not able to produce a strong enough signal for de novo methylation of an unmethylated PAI2 target (Fig 7). The invpai1-pai4 mutant thus illustrates that when superimposed on systems that maintain preexisting methylation imprints, even a severely impaired RNA signal for DNA methylation can promote stable methylation patterning.

	FOOTNOTES

Sequence data from this article have been deposited with the EMBL/GenBank Data Libraries under accession no. AY357734.
¹ Present address: Mayo Clinic, 4500 San Pablo Rd., Jacksonville, FL 32224.

	ACKNOWLEDGMENTS

This work was supported by National Institutes of Health grant GM-61148 and March of Dimes grant FY00-655 to J.B. and by National Institutes of Environmental Health Sciences Training Grant ES 07141 to S.M.

Manuscript received August 6, 2003; Accepted for publication September 24, 2003.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

BARTEE, L. and J. BENDER, 2001  Two Arabidopsis methylation-deficiency mutations confer only partial effects on a methylated endogenous gene family. Nucleic Acids Res. 29:2127-2134.[Abstract/Free Full Text]

BARTEE, L., F. MALAGNAC, and J. BENDER, 2001  Arabidopsis cmt3 chromomethylase mutations block non-CG methylation and silencing of an endogenous gene. Genes Dev. 15:1753-1758.[Abstract/Free Full Text]

BÉCLIN, C., S. BOUTET, P. WATERHOUSE, and H. VAUCHERET, 2002  A branched pathway for transgene-induced RNA silencing in plants. Curr. Biol. 12:684-688.[CrossRef][Medline]

BENDER, J. and G. R. FINK, 1995  Epigenetic control of an endogenous gene family is revealed by a novel blue fluorescent mutant of Arabidopsis. Cell 83:725-734.[CrossRef][Medline]

BIRD, A., 2002  DNA methylation patterns and epigenetic memory. Genes Dev. 16:6-21.[Free Full Text]

DALMAY, T., A. HAMILTON, S. RUDD, S. ANGELL, and D. C. BAULCOMBE, 2000  An RNA-dependent RNA polymerase gene in Arabidopsis is required for posttranscriptional gene silencing mediated by a transgene but not by a virus. Cell 101:543-553.[CrossRef][Medline]

DALMAY, T., R. HORSEFIELD, T. H. BRAUNSTEIN, and D. C. BAULCOMBE, 2001  SDE3 encodes an RNA helicase required for post-transcriptional gene silencing in Arabidopsis.. EMBO J. 20:2069-2077.[CrossRef][Medline]

FAGARD, M., S. BOUTET, J.-B. MOREL, C. BELLINI, and H. VAUCHERET, 2000  AGO1, QDE-2, and RDE-1 are related proteins required for post-transcriptonal gene silencing in plants, quelling in fungi, and RNA interference in animals. Proc. Natl. Acad. Sci. USA 97:11650-11654.[Abstract/Free Full Text]

FINNEGAN, E. J., R. MARGIS, and P. M. WATERHOUSE, 2003  Posttranscriptional gene silencing is not compromised in the Arabidopsis CARPEL FACTORY (DICER-LIKE1) mutant, a homolog of Dicer-1 from Drosophila.. Curr. Biol. 13:236-240.[CrossRef][Medline]

HAUGHN, G. W. and C. SOMERVILLE, 1986  Sulfonylurea-resistant mutants in Arabidopsis thaliana.. Mol. Gen. Genet. 204:430-434.[CrossRef]

JEDDELOH, J. A., J. BENDER, and E. J. RICHARDS, 1998  The DNA methylation locus DDM1 is required for maintenance of gene silencing in Arabidopsis.. Genes Dev. 12:1714-1725.[Abstract/Free Full Text]

KATO, M., A. MIURA, J. BENDER, S. E. JACOBSEN, and T. KAKUTANI, 2003  Role of CG and non-CG methylation in immobilization of transposons in Arabidopsis.. Curr. Biol. 13:421-426.[CrossRef][Medline]

LUFF, B., L. PAWLOWSKI, and J. BENDER, 1999  An inverted repeat triggers cytosine methylation of identical sequences in Arabidopsis.. Mol. Cell 3:505-511.[CrossRef][Medline]

MALAGNAC, F., L. BARTEE, and J. BENDER, 2002  An Arabidopsis SET domain protein required for maintenance but not establishment of DNA methylation. EMBO J. 21:6842-6852.[CrossRef][Medline]

MATZKE, M., A. J. M. MATZKE, and J. M. KOOTER, 2001  RNA: guiding gene silencing. Science 293:1080-1083.[Abstract/Free Full Text]

MELQUIST, S. and J. BENDER, 2003  Transcription from an upstream promoter controls methylation signaling from an inverted repeat of endogenous genes in Arabidopsis.. Genes Dev. 17:2036-2047.[Abstract/Free Full Text]

MELQUIST, S., B. LUFF, and J. BENDER, 1999  Arabidopsis PAI gene arrangements, cytosine methylation and expression. Genetics 153:401-413.[Abstract/Free Full Text]

METTE, M. F., W. AUFSATZ, J. VAN DER WINDEN, M. A. MATZKE, and A. J. M. MATZKE, 2000  Transcriptional silencing and promoter methylation triggered by double-stranded RNA. EMBO J. 19:5194-5201.[CrossRef][Medline]

MIURA, A., S. YONEBAYASHI, K. WATANABE, T. TOYAMA, and H. SHIMADA et al., 2001  Mobilization of transposons by a mutation abolishing full DNA methylation in Arabidopsis.. Nature 411:212-214.[CrossRef][Medline]

MOURRAIN, P., C. BÉCLIN, T. ELMAYAN, F. FEUERBACH, and C. GODON et al., 2000  Arabidopsis SGS2 and SGS3 genes are required for posttranscriptional gene silencing and natural virus resistance. Cell 101:533-542.[CrossRef][Medline]

PÉLISSIER, T., S. THALMEIR, D. KEMPE, H.-L. SÄNGER, and M. WASSENEGGER, 1999  Heavy de novo methylation at symmetrical and non-symmetrical sites is a hallmark of RNA-directed DNA methylation. Nucleic Acids Res. 27:1625-1634.[Abstract/Free Full Text]

SCHRAMKE, V. and R. ALLSHIRE, 2003  Hairpin RNAs and retrotransposon LTRs effect RNAi and chromatin-based gene silencing. Science 301:1069-1074.[Abstract/Free Full Text]

SIJEN, T., I. VIJN, A. REBOCHO, R. VAN BLOKLAND, and D. ROELOFS et al., 2001  Transcriptional and posttranscriptional gene silencing are mechanistically related. Curr. Biol. 11:436-440.[CrossRef][Medline]

SINGER, T., C. YORDAN, and R. A. MARTIENSSEN, 2001  Robertson's Mutator transposons in A. thaliana are regulated by the chromatin-remodeling gene Decrease in DNA Methylation (DDM1). Genes Dev. 15:591-602.[Abstract/Free Full Text]

VAISTIJ, F. E., L. JONES, and D. C. BAULCOMBE, 2002  Spreading of RNA targeting and DNA methylation in RNA silencing requires transcription of the target gene and a putative RNA-dependent RNA polymerase. Plant Cell 14:857-867.[Abstract/Free Full Text]

VIELLE-CALZADA, J.-P., R. BASKAR, and U. GROSSNIKLAUS, 2000  Delayed activation of the paternal genome during seed development. Nature 404:91-94.[CrossRef][Medline]

VOLPE, T. A., C. KIDNER, I. M. HALL, G. TENG, and S. I. S. GREWAL et al., 2002  Regulation of heterochromatic silencing and histone H3 lysine-9 methylation