By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Studies have identified a sub-group of SGS3-LIKE proteins including FDM1–5 and IDN2 as key components of RNA-directed DNA methylation pathway (RdDM). Although FDM1 and IDN2 bind RNAs with 5′ overhangs, their functions in the RdDM pathway remain to be examined. Here we show that FDM1 interacts with itself and with IDN2. Gel filtration suggests that FDM1 may exist as a homodimer in a heterotetramer complex in vivo. The XH domain of FDM1 mediates the FDM1–FDM1 and FDM1–IDN2 interactions. Deletion of the XH domain disrupts FDM1 complex formation and results in loss-of-function of FDM1. These results demonstrate that XH domain-mediated complex formation of FDM1 is required for its function in RdDM. In addition, FDM1 binds unmethylated but not methylated DNAs through its coiled-coil domain. RNAs with 5′ overhangs does not compete with DNA for binding by FDM1, indicating that FDM1 may bind DNA and RNA simultaneously. These results provide insight into how FDM1 functions in RdDM.
In this study, we report that FDM1 acts as a complex in RdDM. FDM1 interacts with both itself and IDN2. Gel filtration analysis suggests that FDM1 exists as a homodimer in a heterotetramer complex that may contain IDN2 in vivo. The FDM1 complex formation depends on its XH domain. Mutant FDM1 protein lacking its XH domain fails to form a complex and is unable to complement the DNA methylation defects of fdm1-1 fdm2-1, demonstrating that XH domain-mediated complex formation by FDM1 is required for its function in RdDM. FDM1 binds DNA in vitro through its coiled-coil domain. RNAs with 5′ overhangs do not abolish the DNA-binding ability of FDM1, indicating that FDM1 may bind both DNA and RNA simultaneously. Through functional analyses of FDM1 protein domains, this study extends our understanding of the RdDM pathway.
FDM1 interacts with itself and IDN2
FDM1 and FDM2 share high identity (approximately 93% identity and 96% similarity). However, the fdm1-1 mutation (null) but not the fdm2-1 mutation (null) individually reduces DNA methylation, indicating that FDM1 may play a major role in RdDM. In addition, expression of FDM1 but not FDM2 in fdm1-1 fdm2-1 is sufficient to restore DNA methylation to wild-type levels. This provides an advantage of being able to study FDM1 function without an effect of FDM2 in vivo by expression of FDM1 mutants. Thus, we focused on FDM1 in this study. In order to obtain insight into how FDM1 acts in RdDM, we tested the interaction of FDM1 with known RdDM components including DRM2, DMS3, RDR2, SPT5L, FDM1 and IDN2 using the pGBKT7/pGADT7 two-hybrid system. In this system, a protein of interest is fused with a DNA-binding domain in the pGBKT7 plasmid, while the potential interactor is fused with a transcriptional activation domain in the pGADT7 vector. If two proteins interact, the DNA-binding domain associates with the transcriptional activation domain after co-transformation into yeast cells. This activates expression of a report gene that produces adenine (Ade), and thus enables growth of the yeast strain in the absence of Ade. Co-transformation of pGADT7-FDM1/pGBKT7-FDM1 and pGADT7-FDM1/pGBKT7-IDN2 pairs enabled growth of yeast cells in the absence of Ade (Figure 1a). In contrast, yeast cells failed to grow on in the absence of Ade after co-transformation of pGADT7/pGBKT7-FDM1, pGADT7/pGBKT7-IDN2 or pGBKT7/pGADT7-FDM1 (Figure 1a). These results indicate that FDM1 may interact with itself and IDN2. This assay did not detect interaction of FDM1 with DRM2, DMS3, RDR2 or SPTL5 (Figure 1b). The FDM1–RDR2 interaction result from this assay is consistent with the pulldown results in Xie et al., (2012).
The XH domain of FDM1 is necessary for FDM1–FDM1 and FDM1–IDN2 interactions
To identify the protein domains of FDM1 responsible for the interaction, we generated a series of truncation mutants of FDM1 in pGADT7 (Figure 2a): lacking the XH domain (FDM1-T1), the XH domain alone (FDM1-T2), lacking the XH and coiled-coil domains (FDM1-T3) and lacking the zinc finger and the XS domain (FDM1-T4). We tested the interaction of these truncated FDM1 mutants with full-length FDM1 and IDN2 using the yeast two-hybrid assay described above. FDM1-T2 and FDM1-T4 were able to interact with FDM1 and IDN2, respectively, because co-transformation of these pairs enabled yeast cell to grow in the absence of Ade (Figure 2b,c). In contrast, FDM1-T1 and FDM-T3 did not interact with FDM1 and IDN2, respectively. These results indicated that the XH domain of FDM1 is necessary for FDM1–FDM1 and FDM1–IDN2 interactions. However, the yeast cells containing FDM1/FDM1-T2 (XH domain alone) and IDN2/FDM1-T2 grew more slowly than those containing FDM1/FDM1 and IDN2/FDM1, respectively (Figures 1 and 2). This result indicated that full-strength FDM1–FDM1 and FDM1–IDN2 interactions may require additional protein domains.
To validate the function of the XH domain in protein–protein interactions, we replaced the tryptophan at position 605 (W605) and the glutamic acid at position 617 (E617) with alanine (A) in the XH domain of FDM1 (FDM1-T5; Figure 2a). These two amino acids are conserved in XH domains and hence play important roles in mediating protein–protein interactions. As shown in Figure 2(d), FDM1-T5 did not interact with FDM1 and IDN2. This result confirmed that the XH domain of FDM1 is necessary for FDM1–FDM1 and FDM1–IDN2 interactions.
FDM1 pulls down FDM1 and IDN2 in vitro
To further confirm FDM1–IDN2 interaction, we performed an in vitro pulldown assay. We expressed recombinant IDN2 protein fused to a maltose-binding protein epitope at its N-terminus (MBP–IDN2), FDM1 fused to an N-terminal GST tag (GST–FDM1) or the controls MBP, GST and GST–FDM1ΔXH (FDM1 lacking the XH domain) in Escherichia coli. After expression, protein extracts containing MBP–IDN2 were mixed with extracts containing GST–FDM1, and reciprocal pulldown was then performed using amylose resin or glutathione beads. To avoid DNA- or RNA-mediated protein interactions, we treated the samples with micrococcal nuclease that digests both DNA and RNA. Enrichment of MBP–IDN2 in the GST–FDM1 complex and of GST–FDM1 in the MBP–IDN2 complex was detected using antibodies against MBP or GST, respectively (Figure 3a,b). In contrast, GST and GST–FDM1ΔXH failed to pull down MBP–IDN2, and MBP did not pull down GST–FDM1 (Figure 3a,b). To validate FDM1–FDM1 interaction, we mixed protein extracts containing YFP–FDM1 or YFP–FDM1ΔXH with extracts containing GST–FDM1 or GST, respectively, and performed a reciprocal pulldown assay. GST–FDM1 and YFP–FDM1 reciprocally pulled down each other (Figure 3c,d), but YFP–FDM1ΔXH and GST did not interact with GST–FDM1 and YFP–FDM1, respectively (Figure 3c,d).
FDM1 forms a tetramer in vitro
The yeast two-hybrid and pulldown analyses suggest that FDM1 interacts with itself through its XH domain. Thus, we examined whether FDM1 forms a dimer or an oligomer complex. We first expressed recombinant FDM1 fused with a C-terminal 6 × His tag (FDM1-His) and a truncated FDM1-His lacking XH (FDM1ΔXH-His) in E. coli, and purified the resulting proteins. The FDM1-His or FDM1ΔXH-His proteins were then analyzed by size-exclusion HPLC. The elution fractions were then separated in SDS–PAGE and probed with antibodies recognizing the His tag. The column was calibrated using protein standards. We obtained information on the relative size of the FDM1 complex by comparing fractions of FDM1 with peak elution times of standard proteins. FDM1-His had a peak elution time of 114–118 min (Figure 4a), suggesting that FDM1-His may exist as a tetramer complex of approximately 300 kDa. In contrast, FDM1ΔXH-His eluted at between 144 and 148 min, corresponding to the size of the FDM1ΔXH monomer (approximately 60 kDa; Figure 4a). These analyses revealed that FDM1 forms a tetramer complex that requires the XH domain for its formation.
Because FDM1 also interacts with IDN2, we next tested whether incubation of IDN2 and FDM1 generates a larger complex or tetramer in order to obtain insight into the nature of the FDM1–IDN2 complex. We purified MBP–IDN2 and removed the MBP tag. However, incubation of IDN2 with FDM1-His still produced a tetramer (Figure 4b). This result indicates that FDM1 and IDN2 may form a tetramer in vitro. However, the copy numbers of FDM1 and IDN2 in the complex remain to be determined.
FDM1 protein exists as a dimer in a tetramer complex in vivo
To obtain information on the FDM1 complex in vivo, we analyzed Arabidopsis protein extracts containing YFP–FDM1 by size-exclusion HPLC. YFP–FDM1 complemented the DNA methylation defects in fdm1-1. Anti-YFP antibody detected the presence of YFP–FDM1 in a complex of approximately 350 kDa as calculated from a standard curve produced using the elution times of protein standards (Figure 4c). The calculated molecular mass for YFP–FDM1 is approximately 100 kDa, and that for untagged FDM proteins and IDN2 is approximately 75 kDa. Thus, the mass of approximately 350 kDa equals the molecular mass of two copies of YFP–FDM1 and two copies of other untagged FDM proteins or IDN2. This result suggests that FDM1 may exist as a homodimer in a heterotetramer complex.
The XH domain is required for the function of FDM1 in RdDM
Next, we examined whether the XH domain was required for FDM1 function in RdDM. We generated transgenic fdm1-1 fdm2-1 plants expressing either 35S::YFP-FDM1 or 35S:YFP-FDM1ΔXH lacking the XH domain. In a previous study, we showed that expression of FDM1 under the direction of its native promoter is sufficient to complement the DNA methylation defects of fdm1-1 fdm2-1 (Xie et al., 2012). Using fdm1-1 fdm2-1 enabled us to test the function of the XH domain of FDM1 without effects of FDM2, which has a 96% similarity with FDM1. The transcript levels of transgenes and their products were similar in all four transgenic lines (Figure 5a,b). We examined the methylation levels of two RdDM targets, AtSN1 and ING5, AtSN1 and IGN5 in two transgenic fdm1-1 fdm2-1 lines harboring 35S::YFP-FDM1 and two transgenic fdm1-1 fdm2-1 lines harboring 35S:YFP-FDM1ΔXH using methylation-sensitive HaeIII restriction enzyme digestion followed by PCR. Less DNA methylation at AtSN1 and IGN5 results in less PCR product after HaeIII digestion because it cuts unmethylated but not methylated DNA. As shown in Figure 5(c), the 35S::YFP-FDM1 transgene restored the DNA methylation content of AtSN1 and IGN5 in fdm1-1 fdm2-1 to wild-type levels. In contrast, the DNA methylation levels of fdm1-1 fdm2-1 harboring 35S:YFP-FDM1ΔXH were comparable with those in fdm1-1 fdm2-1. Consistent with this, silencing of AtSN1 transcription was not restored in fdm1-1 fdm2-1 harboring 35S:YFP-FDM1ΔXH (Figure 5d). These results demonstrate that the XH domain is essential for the function of FDM1 in RdDM.
FDM1 binds unmethylated but not methylated DNA
Protein sequence analyses showed that the coiled-coil domain of FDM1 has approximately 50% similarity to a portion of the SMC (structural maintenance of chromosomes) protein from Methanocaldococcus sp. FS406-22 (Figure 6a). As SMC proteins bind DNAs, this finding prompted us to test whether FDM1 binds DNA using a GST pulldown assay. This method reduces the background signal because it eliminates unbound probes. This method has been used previously to study protein–nucleic acid interactions (Jiao et al., 2002; Yu et al., 2008). We incubated purified GST–FDM1 with a 50 bp 32P-labeled DNA fragment and a 50 nucleotide 32P-labeled single-stranded DNA (ssDNA) that corresponds to a fragment of AtSN1 DNA (Figure 6b). After washing, the DNAs were extracted from beads and separated on a native PAGE gel. GST–FDM1 but not GST alone retained the 50 bp DNA fragment (Figure 6b). However, FDM1 was unable to bind the ssDNA (Figure 6b). Addition of unlabeled DNA with the same sequences eliminated the radioactive signals. These results indicated that FDM1 binds DNA (Figure 6b). FDM1 also bound a DNA fragment containing a poly(A) strand and a poly(T) strand (Figure 6d). This result suggests that DNA binding by FDM1 is not sequence-specific. However, FDM1 did not bind methylated DNA (Figure 6d). To identify protein domains required for the DNA-binding ability of FDM1, we expressed and purified a series of truncated FDM1 proteins fused to an N-terminal GST tag (Figure 6e,f). The truncated FDM1 protein lacking a portion of the coiled-coil domain but not other domains failed to bind DNA (Figure 6f). In addition, the coiled-coil domain itself was able to bind DNA (Figure 6f). Based on these results, we propose that the coiled-coil domain is necessary and sufficient for DNA binding of FDM1.
We have shown that FDM1 binds RNA with 5′ overhangs, and this depends on the XS domain. This raised a question of whether FDM1 can bind DNA and RNA simultaneously. To address this question, we examined whether addition of unlabeled RNAs with 5′ overhangs affects the DNA-binding ability of FDM1. If FDM1 binds DNA and RNA at the same time, addition of RNAs should not eliminate DNA binding of FDM1. As shown in Figure 6(c), addition of RNAs with 5′ overhangs did not affect DNA retention of FDM1.
Studies on FDM1 and IDN2 have suggested that they may act in downstream of RdDM, presumably by stabilizing the duplex of ra-siRNA–Pol V-dependent transcripts (Ausin et al., 2009; Zheng et al., 2010; Xie et al., 2012). In this study, we demonstrate that FDM1 exists in a complex for proper function in RdDM, and is an RNA- and DNA-binding protein.
Yeast two-hybrid and in vitro protein pulldown experiments showed that FDM1 interacts with IDN2. Given its high similarity with FDM1, FDM2 most likely also interacts with IDN2. While this paper was in preparation, two other groups found that IDN2 complexes contain IDN2 PARALOG 1 (IDP1)/IDN2-LIKE1 (IDNL1) and IDP2/IDNL2 (Ausin et al., 2012; Zhang et al., 2012). IDP1/IDNL1 and IDP2/IDNL2 are synonymous with FDM1 and FDM2, respectively. These results demonstrate that FDM1/IDP1/IDNL1 and FDM2/IDP2/IDNL2 form a complex with IDN2. We detected a FDM1/IDP1–FDM1/IDP1 interaction in a yeast two-hybrid assay, but Zhang et al. (2012) did not. This discrepancy may be due to the fact that different yeast strains were used (PJ694A versus AH109). Protein pulldown (Figure 3) and gel filtration (Figure 4) experiments further confirmed the FDM1–FDM1 interaction. FDM1 forms a homotetramer in vitro, but may exist as a homodimer in a tetramer complex in vivo (Figure 4). Multidimensional protein identification technology (MudPIT) analysis showed that IDN2 may be the only partner of IDNL1/FDM1 (Ausin et al., 2012). Crystal structure and yeast two-hybrid analyses revealed that IDN2 lacking the XH domain forms a homodimer in vitro (Ausin et al., 2012; Zhang et al., 2012). These results suggest that FDM1 and IDN2 form a heterotetramer containing an FDM1 dimer and an IDN2 dimer. FDM2 is present in the IDN2 complex and is highly similar to FDM1, indicating the presence of an IDN2–IDN2–FDM2–FDM2 tetramer. The presence of these two complexes is consistent with the functional redundancy of FDM1 and FDM2 (Xie et al., 2012). However, it is possible that IDN2–IDN2–FDM1–FDM2 is present in low amounts such that MudPIT analysis cannot detect IDNL2/FDM2 in the IDNL1/FDM1 complex.
Whole-genome bisulfite sequence analysis revealed that DNA methylation patterns are similar in idn2-1, idnl1-1 idnl2-1 (fdm1 fdm2) and idn2-1 idnl1-1 idnl2-1, indicating that IDN2, FDM1/IDNL1 and FDM2/IDNL2 mostly likely function together (Ausin et al., 2012). IDN2, FDM1/IDNL1 and FDM2/IDNL2 affect most DRM2 targets and few non-DRM2 targets, indicating that they mainly act in the RdDM pathway (Ausin et al., 2012). The DNA methylation defect in idn2-1, idnl1-1 idnl2-1 and idn2-1 idnl1-1 idnl2-1 is weaker than that in drm2. This may be due to the redundant functions of homologs of IDN2, FDM1 and FDM2. Indeed, three FDM1 homologs, FDM3, FDM4 and FDM5, act redundantly with FDM1 (Xie et al., 2012). Of these, FDM3 and FDM4 are in the IDN2 sub-family, whereas FDM5 is grouped with FDM1 and FDM2. The IDN2 complex does not contain FDM3, FDM4 and FDM5 (Ausin et al., 2012; Zhang et al., 2012), raising the possibility that other FDM complexes may exist. Perhaps the IDN2/FDM1 (FDM2) complex plays a major role in RdDM, while others have minor functions, because loss of function of either IDN2 or FDM1 alone causes a DNA methylation defect but loss of function of other FDM proteins alone does not (Xie et al., 2012). Alternatively, they may play different roles to the IDN2/FDM1 complex.
The function of the XH domain was previously unknown. We found that FDM1 protein lacking the XH domain or harboring mutations in its XH domain failed to interact with itself or with IDN2 (Figures 2–4). The XH domain of FDM1 by itself interacts with FDM1 and IDN2, demonstrating that the XH domain of FDM1 functions in mediating protein–protein interaction. In addition, IDN2 without a functional XH domain fails to interact with IDP1/FDM1 (Zhang et al., 2012). Both FDM1 and IDN2 lacking the XH domain failed to rescue DNA methylation defects in the corresponding mutants (Figure 5) (Zhang et al., 2012), demonstrating that XH domain-mediated complex formation is essential for their function in RdDM.
It has been shown that FDM1/IDNL1 or FDM2/IDNL2 cannot replace IDN2 in their complexes because a strong idn2-1 allele has similar DNA methylation defects to idnl-1 idnl-2 and idn2-1 idnl1-1 idnl2-1 (Ausin et al., 2012). However, a weak idn2-3 mutation reduced DNA methylation in fdm1-1 and fdm2-1, respectively (Xie et al., 2012), indicating the function of the IDN2/FDM1/FDM2 complex is further impaired in the double mutants. What causes the difference between FDM1 and IDN2? For both FDM1 and IDN2, the XH domain mediates protein–protein interaction and the XS domain binds dsRNAs with 5′ overhangs, indicating that they may not be the reasons for the differences between FDM1 and IDN2. We found that the coiled-coil domain of FDM1 binds DNA and is not required for FDM1–FDM1 and FDM1–IDN2 interactions. In contrast, the coiled-coil domain of IDN2 has been shown to mediate IDN2–IDN2 interaction (Ausin et al., 2012; Zhang et al., 2012). Thus, the coiled-coil domain of FDM1 is biochemically different from that of IDN2, and may be the factor that distinguishes FDM1 from IDN2. Given the high similarity between FDM1 and FDM2, this most likely is the cause of the difference between FDM2 and IDN2 as well.
FDM1 and IDN2 bind RNAs with 5′ overhangs through their XS domains (Ausin et al., 2009; Xie et al., 2012; Zhang et al., 2012). In addition, FDM1 binds DNA in a non-sequence-specific manner through the coiled-coil domain, and DNA binding of FDM1 is not be competed for by RNA, indicating that FDM1 may bind DNA and RNA simultaneously. These results have advanced the model for IDN2/FDM1 function (Ausin et al., 2009, 2012; Zheng et al., 2010; Xie et al., 2012). The XS domain of FDM1 (FDM2) and IDN2 may bind the duplex of AGO4-bound ra-siRNA and Pol V-dependent transcripts (Ausin et al., 2009, 2012; Zheng et al., 2010; Xie et al., 2012). This binding will recruit the FDM1–IDN2 or FDM2–IDN2 complex to RdDM loci. Subsequently, the coiled-coil domain of FDM1 (FDM2) binds the DNA. Binding of the FDM1–IDN2 complex to the RNA duplex and RdDM target loci may have two roles that are not mutually exclusive. One is to prevent potential cleavage of Pol V-dependent transcripts by the AGO4–ra-siRNA complex, which may disrupt the AGO4–chromatin interaction. However, the levels of Pol V transcripts are not affected by fdm1 fdm2 and idn2 mutations (Ausin et al., 2012; Xie et al., 2012), arguing against this possibility. The other is that the FDM1 complex may provide a marker for DRM2 to recognize. However, FDM1 does not bind methylated DNA, indicating that the FDM1 complex may be required for initiation but not reinforcement of DNA methylation. The yeast two-hybrid assay does not identify the FDM1–DRM2 interaction, suggesting that other factors may be involved. Clearly this model needs to be examined using an FDM1 mutant deficient in DNA and/or RNA binding.
Plant materials and growth conditions
Plants were grown at 22 °C under long-day conditions (16 h light/8 h dark). fdm1-1 (SALK_075813), fdm2-1 (SAIL_291_F01) and fdm1-1 fdm2-1 are in the Columbia genetic background (Xie et al., 2012).
YFP cDNA was cloned into binary vector pMDC32 (Curtis and Grossniklaus, 2003) to generate pMDC32-YFP. Then FDM1 and FDM1ΔXH (lacking the XH domain) cDNAs were PCR-amplified and cloned into pMDC32-YFP to generate p35S::FDM1-YFP and p35S::FDM1ΔXH-YFP constructs, respectively. The FDM1, truncated FDM1, IDN2, RDR2, DRM2, SPTL5 and DMS3 cDNAs were PCR-amplified and cloned into the pGADT7 and/or pGBKT7 (Clontech, http://www.clontech.com/) vectors to produce constructions used for the yeast two-hybrid assay. The full-length FDM1 and truncated FDM1 cDNAs were PCR-amplified and cloned into pGEX-2TK(GE Healthcare, http://www3.gehealthcare.com/) or pET28(a), (EMD Millipore, http://www.emdmillipore.com/chemicals) vectors to generate GST or 6 × His fusion constructions. The IDN2 cDNA was PCR-amplified and cloned into the pMAL-c5X(NEB, http://www.neb.com/nebecomm/default.asp) vector to generate the MBP–IDN2 fusion construct. The primers used for plasmid constructions are listed in Table S1.
p35S::FDM1-YFP and p35S::FDM1ΔXH-YFP were transformed into fdm1-1 or fdm1-1 fdm2-1, respectively. The T1 transgenic plants were selected by hygromycin resistance.
Yeast two-hybrid assay
Various plasmid pairs were co-transformed into yeast strain AH109. Selective drop-out medium lacking leucine and tryptophan was used to select yeast containing the plasmid pairs. The resulting clones were diluted in 50 μl water, and 5 μl was used for spot assays on selective drop-out medium lacking leucine, tryptophan and adenine. FDM1–FDM1 and FDM1–IDN2 interactions activate expression of Ade, which enables the growth of AH109 cells in the absence of Ade.
GST-, MBP- or His-tagged proteins were expressed in E. coli. BL21 and extracted as described previously (Xie et al., 2012). The YFP–FDM1 and YFP–FDM1ΔXH constructs were transiently expressed in tobacco (Nicotiana Benthamiana) and extracted as described by Yu et al. (2008).
Protein pulldown assay
Protein extracts containing GST, GST–FDM1 or GST–FDM1ΔXH were mixed with equal volumes of protein extracts containing MBP–IDN2, YFP and YFP–FDM1, respectively. The mixed lysate was incubated with antibodies against GFP (and GFP variants) coupled to protein A agarose beads (Clontech), amylose resin (NEB) or glutathione Sepharose 4B beads (GE Healthcare) for 4 h.
The precipitates were washed five times (5min for each wash) with extraction buffer, separated by SDS–PAGE and blotted using antibodies recognizing the MBP, GST or YFP tag.
FDM1-His and FDM1ΔXH-His were purified using Ni-resin (EMD Millipore) according to the manufacturer’s instruction. After elution from Ni-resin, 100 μl protein solution was passed through a 0.22 μm filter and loaded onto the column. The gel filtration was performed on an HPLC system and a HiPrep 16/60 Sephacryl S-300 HR column (GE Healthcare) at a rate of 0.5 ml/min, and 0.5 ml fractions were collected every minute. For gel filtration of Arabidopsis protein extracts, collected fractions were precipitated with acetone at −20 °C overnight and resuspended in SDS loading buffer. Fractions were separated by 8% SDS–PAGE and analyzed by Western blotting using antibodies recognizing His or YFP. The protein standards (Bio-Rad, http://www.bio-rad.com/) used to calibrate the column contain five size standards, and the elution times for each peak are 94 min for 670 kDa, 129 min for 158 kDa, 150 min for 44 kDa, 173 min for 17 kDa, and 233 min for 1.35 kDa.
DNA methylation and RT-PCR analysis
The DNA methylation assay was performed as described previously (Xie et al., 2012). Genomic DNAs extracted from flowers were digested with HaeIII, and 5% of the digested DNA was used for PCR amplification of AtSN1 and IGN5. Simultaneously, undigested genomic DNA was amplified as a quantity control. After DNase I treatment, 5 μg total RNAs from inflorescences were used to synthesize cDNA with SuperScript III (Invitrogen, http://www.invitrogen.com/) and oligo(dT) primers. The diluted cDNA was used to amplify AtSN1 by PCR. Amplification of UBQ5 was used as a loading control.
GST–FDM1 and GST-tagged FDM1 mutants were purified as described by Xie et al. (2012). A 5′ overhanging dsRNA probe was generated as described previously (Xie et al., 2012). A 50 nucleotide single-strand DNA fragment corresponding to a portion of AtSN1 DNA was labeled at its 5′ end using T4 polynucleotide kinase (NEB) in the presence of [α32P]ATP. Annealing this ssDNA to its complementary strand produced double-stranded DNA. The DNA- and RNA-binding assays were performed as previously described (Jiao et al., 2002). Methylated DNA and its unmethylated control were synthesized by Integrated DNA Technologies (http://eu.idtdna.com/site) as described previously (Ito et al., 2003).
We thank Drs Wayne R. Riekhof, Xiang Liu and Shuxin Zhang School of Biological Sciences, University of Nebraska-Lincoln, NE, USA for critical reading of the manuscript. This work was supported by the National Science Foundation (grant number MCB-1121193 to B.Y.).