Letter | Published:

Telobox motifs recruit CLF/SWN–PRC2 for H3K27me3 deposition via TRB factors in Arabidopsis

Nature Geneticsvolume 50pages638644 (2018) | Download Citation


Polycomb repressive complexes (PRCs) control organismic development in higher eukaryotes through epigenetic gene repression1,2,3,4. PRC proteins do not contain DNA-binding domains, thus prompting questions regarding how PRCs find their target loci5. Here we present genome-wide evidence of PRC2 recruitment by telomere-repeat-binding factors (TRBs) through telobox-related motifs in Arabidopsis. A triple trb1-2, trb2-1, and trb3-2 (trb1/2/3) mutant with a developmental phenotype and a transcriptome strikingly similar to those of strong PRC2 mutants showed redistribution of trimethyl histone H3 Lys27 (H3K27me3) marks and lower H3K27me3 levels, which were correlated with derepression of TRB1-target genes. TRB1–3 physically interacted with the PRC2 proteins CLF and SWN. A SEP3 reporter gene with a telobox mutation showed ectopic expression, which was correlated with H3K27me3 depletion, whereas tethering TRB1 to the mutated cis element partially restored repression. We propose that telobox-related motifs recruit PRC2 through the interaction between TRBs and CLF/SWN, a mechanism essential for H3K27me3 deposition at a subset of target genes.


In Drosophila, Polycomb response elements (PREs) containing multiple binding sites for several transcription factors are predominantly involved in the recruitment of PRC2 or PRC1, which catalyze formation of H3K27me3 and monoubiquitinated histone H2A Lys119 (ref. 5), respectively. The recruitment of either PRC1 or PRC2 appears to be sufficient to target both complexes to a locus6,7,8,9. In mammalian cells, recruiter proteins and cognate motifs have been proposed, as well as PRC1 and PRC2 recruitment through long noncoding RNAs and signals associated with unmethylated CpG islands5,10. In Arabidopsis, analyses of PRC-target genes have identified several cis elements with PRE-like properties3. An intronic RY motif is required for stable silencing of FLC after vernalization11,12. The RLE element, which contains GAGA and RY motifs, prevents ectopic expression of LEAFY COTYLEDON 2 (LEC2)13. A complex cis element at KNOX homeobox genes binds the ASYMMETRIC LEAVES (AS) repressor complex, which interacts with PRC2 and restricts expression of target genes to stem cell niches14. At a genome-wide scale, B3-domain transcription factors related to VAL1 and VAL2 have been suggested to be widely involved in PRC1 recruitment through RY motifs15,16. Two transcription factors, BASIC PENTACYSTEINE (BPC) 1 and ARABIDOPSIS ZINC FINGER1 (AZF1), recruit the PRC2 component FERTILIZATION INDEPENDENT ENDOSPERM (FIE) to target genes containing both GAGA motifs and teloboxes17. Furthermore, the GAGA-motif-binding protein BPC6 interacts with the PRC component LIKE HETEROCHROMATIN PROTEIN 1 (LHP1)18.

As we previously reported for lhp1 (also known as tfl2) (ref. 19), the phenotype of the PRC2 mutant clf was enhanced by a combination with loss-of-function alleles for trb1 and/or trb3 (Supplementary Fig. 1a–c), whereas a trb2-1 loss-of-function allele did not enhance the clf phenotype (Supplementary Fig. 1d,e). Although none of the trb single or double mutants appeared to be different from wild-type controls (Supplementary Figs. 1 and 2), the trb1/2/3 triple mutant was severely affected, showing curling of cotyledons early in development and slow growth of roots and leaves (Fig. 1a,b). A genomic TRB2 fragment that included a C-terminal Flag epitope sequence complemented the trb1/2/3 phenotype (Supplementary Fig. 2b). A second TRB2 loss-of-function allele (trb2-2), generated by CRISPR–CAS9 editing20 in the trb1/3 background, reproduced the trb1/2/3 phenotype (Supplementary Fig. 2c,d).

Fig. 1: The phenotype of trb1/2/3 mutants is similar to those of strong PRC mutants.
Fig. 1

a, Leaf phenotype of 10-d-old seedlings of trb1/2/3, trb1/3 trb2–/+, and Col-0, as indicated. b, Whole 10-d-old seedlings of Col-0, trb1/2/3, and trb1/3 trb2–/+. c, Heat map showing hierarchical clustering of differentially expressed genes. Red and yellow represent up- and downregulation in mutants, respectively. d, Venn diagrams showing the overlap between up- or downregulated genes in trb1/2/3 and clf-29/swn-21. P values were calculated with two-tailed hypergeometric tests. e, Scatter plots showing a positive correlation between up- and downregulated genes in trb1/2/3 and clf-29/swn-21 mutants. WT, wild type.

In general, TRB2 colocalizes with TRB1 and TRB3 in tobacco cells and Arabidopsis plants19,21,22 (Supplementary Fig. 3a–d). To evaluate whether telomere defects caused the trb1/2/3 phenotype, we performed terminal restriction fragment analysis in trb1/2/3 triple mutants and their trb1/3 trb2-1–/+ siblings. Both genotypes showed shorter telomere length than that in Col-0, and no differences were observed between the second and third generations of propagation in the trb1/3 trb2-1–/+ background (Supplementary Fig. 3e). In contrast to trb1/2/3 seedlings, trb1/3 trb2-1–/+ seedlings grew similarly to wild-type seedlings (Fig. 1b). Thus, the lower dosage of TRBs in trb1/3 trb2-1–/+ mutants was sufficient to cause telomere shortening, which was unlinked to the developmental phenotype of the trb1/2/3 triple mutant.

We investigated the molecular phenotype of the trb1/2/3 mutant through RNA-seq transcriptome profiling. Because previous genetic analyses have suggested that TRBs aid in LHP1 and PRC2 function by maintaining gene repression19 (Supplementary Fig. 1), the transcriptome of trb1/2/3 was compared with those of PRC mutants3,6,23. Arabidopsis encodes 12 homologs of the four canonical PRC2 components and five homologs of the two PRC1 RING-domain components, BMI1 and RING1 (refs 2,6,7,24). Combining mutant alleles of partially redundant genes allows for gradients of relatively weak to strong PRC mutants to be compared. The strongest viable PRC2 mutants were affected in clf and the partially redundant paralog swinger (swn), which encode the catalytic core of PRC2, and were also affected in short interfering RNA–knockdown lines (denoted siFIE) targeting the single-copy gene FIE1. The transcriptome profile observed in trb1/2/3 was closest to those of the strongest PRC mutants (Fig. 1c and Supplementary Dataset 1). Among 990 downregulated and 1,558 upregulated genes in trb1/2/3, a substantial number, 179 and 359, respectively, shared the same direction of misregulation in clf/swn double mutants (Fig. 1d). In contrast, the transcriptome profiles of lhp1, trb1, and trb3 single mutants, as well as their combinations, clustered separately (Fig. 1c). Gene Ontology terms related to plant hormone responses and development were overrepresented for trb1/2/3 upregulated genes, whereas the downregulated genes related to metabolic pathways (Supplementary Dataset 2). Key genes involved in controlling flowering time, such as SEP3 and AGAMOUS (AG), were upregulated in trb1/2/3 (Supplementary Dataset 1)25. Moreover, changes in the trb1/2/3 and clf/swn mutants were positively correlated (Fig. 1e), and commonly upregulated genes were often marked by H3K27me3 (Supplementary Fig. 4).

Because the transcriptome of trb1/2/3 was most similar to those of severe PRC2 mutants, we probed trb1/2/3 seedlings for altered H3K27me3 levels. Most H3K27me3-enriched regions were common between Col-0 and trb1/2/3, but 609 and 730 regions had respectively lower and higher H3K27me3 coverage in Col-0 than in trb1/2/3, corresponding to 22% of all H3K27me3-positive regions (Fig. 2a). These regions overlapped with 982 and 762 genes that significantly lost (Col-0 biased genes) or gained H3K27me3 (trb1/2/3 biased genes) (Fig. 2b and Supplementary Dataset 3).

Fig. 2: Genome-wide changes of H3K27me3 in trb1/2/3 mutants.
Fig. 2

a, MA plot of all H3K27me3-positive peaks from Col-0 and trb1/2/3 seedlings after analysis with MAnorm35. The x axis represents the A value of each peak as average intensity. The y axis represents the M value as log2(difference in intensity). Positive and negative M values indicate higher and lower H3K27me3 levels, respectively, in Col-0 than in trb1/2/3. The color range represents –log10 (P values, two-tailed test) associated with normalized peaks. Two independently grown and processed replicates were analyzed with ChIP–seq. b, Heat map of ChIP–seq read density for gene groups categorized as trb1/2/3 biased, Col-0 biased, and common. Genes aligned at the transcription start site (TSS) are plotted with 5 kb up- and downstream regions. c, Enrichment of trb1/2/3 up- and downregulated genes per H3K27me3 category. Each category was compared with the expected random overlap (dotted line); the x axis represents enrichment score. A two-tailed Fisher’s exact test was used to test the significance. d, IGV screen shot showing a quantitative decrease in H3K27me3 in trb1/2/3 compared with Col-0. e, IGV screen shot showing a quantitative induction of H3K27me3 in trb1/2/3 compared with Col-0. f, IGV screen shot showing no significant changes in H3K27me3 in trb1/2/3 compared with Col-0. Blue boxes indicate the positions of the teloboxes. Chr, chromosome.

Of the 1,744 genes with altered H3K27me3 levels, 300 were differentially expressed (Supplementary Dataset 1). Changes in expression and H3K27me3 coverage showed a negative correlation (Supplementary Fig. 9), thus resulting in a significant enrichment in Col-0 biased and trb1/2/3 biased genes among the 1,558 transcriptionally upregulated and 990 downregulated genes, respectively (Fig. 2c). Gene Ontology–term analysis of Col-0 biased genes identified an enrichment in transcription factors, whereas trb1/2/3 biased genes were associated with metabolic-pathway genes and transcription factors (Supplementary Dataset 4). AG (AT4G18960), AT3G46780, and AT1G05800 are typical examples of Col-0 biased, common, and trb1/2/3 biased genes, respectively (Fig. 2d–f and Supplementary Fig. 5). Together, our results indicated that a substantial redistribution of H3K27me3 in trb1/2/3 mutants compared with Col-0 had a strong effect on transcription.

TRB1 chromatin immunoprecipitation (ChIP)–seq data from wild-type seedlings26 were probed to determine whether changes in H3K27me3 enrichment were direct effects of trb1/2/3 mutation (Supplementary Dataset 5). Col-0 biased genes, compared with trb1/2/3 biased genes, showed robust binding of TRB1 (Fig. 3a). Overall, the enrichment of TRB1 at target genes was significantly higher for Col-0 biased genes than for common H3K27me3-target genes, whereas trb1/2/3 biased genes were less enriched (Fig. 3b). We compared binding of the PRC2 component FIE across all gene groups, using an available dataset from wild-type seedlings27 (Supplementary Dataset 5). Regions bound by TRB1 and FIE overlapped significantly with H3K27me3 and with each other (Fig. 3c). Although the enrichment score for H3K27me3 was higher for regions bound by both TRB1 and FIE than for regions bound by either alone, none of the states appeared to fully depend on the others (Fig. 3d).

Fig. 3: Lower H3K27me3 correlates with TRB1 binding and telobox enrichment.
Fig. 3

a, Heat map of TRB1 (ref. 19) and FIE27 ChIP–seq signals in wild-type seedlings at regions 5 kb up- and downstream of the TSS for Col-0 or trb1/2/3 biased genes19,30. b, Enrichment scores of TRB1 targets among trb1/2/3 biased (1), common (2), and Col-0 biased (3) genes. A two-tailed Fisher’s exact test was used to test the significance. c, Mutual overlap among TRB1, H3K27me3, FIE, and TRB1 + FIE peaks. Significance was tested by permutation (n = 100)36. Heat map shows enrichment scores \(Z={{\rm{log}}}_{2}\left(\frac{{\rm{observed}}\,{\rm{overlap}}}{{\rm{expected}}\,{\rm{overlap}}}\right)\) (−log10(P value)). d, Overlap as in c, expressed as a percentage. e, Motifs enriched at the peak center (±2 kb) of H3K27me3-target regions categorized as Col-0 biased, trb1/2/3 biased, or common. Telobox and Z box were collected from AGRIS37; P0510F09.23, NAC043, FUS3, SPL11, and MEF2B were collected from JASPAR38; SEF1, RY repeat, TATA box, HDZIP2ATATHB2, GARE, and ACGT motif were collected from PLACE39; TBP and TCP16 were collected from AthaMap40. f, Frequency of telobox (blue) and RY (red) motifs at the center (±2 kb) of H3K27me3 peaks, plotted per M value. Lines represent mean values; clouds represent ± s.e.m (= 21/135/574/2,465/2,344/474/104/31 for 8 consecutive equidistant bins). g, FIE binding at Col-0 biased loci (SEP3, EMF1, and AG) in transfected protoplasts prepared from Col-0 or trb1/2/3 seedlings. Plots show data from three independently transfected protoplast preparations. ACTIN2 (ACT2) is the H3K27me3 negative control. Vertical central line and whiskers indicate mean ± s.d. Statistical analysis was performed with two-tailed Student’s t tests (degrees of freedom (df) = 4); NS, not significant. h, TRB1 binding at regions as in g in transfected protoplasts prepared from Col-0 or clf-28 seedlings. Data were generated, plotted and analyzed as in g.

We performed a cis-motif enrichment analysis for Col-0 biased, common, and trb1/2/3 biased genes to uncover factors in addition to TRB1 and FIE binding that define the chromatin state. On the basis of previous ChIP–seq and in vitro binding data, TRB1 binds telobox (AAACCCTA) and telobox-related (RMCCTA)19,28 motifs19,29. Furthermore, telobox-related motifs are enriched at FIE-binding sites27, particularly at genes that are upregulated in lhp1 and clf mutants23. In accordance with the observed TRB1 binding, telobox and telobox-related motifs were overrepresented in Col-0 biased fragments and depleted in trb1/2/3 biased fragments (Fig. 3e and Supplementary Fig. 6). (G)CATGC motifs (also named RY-, FUS3-, and LEC2-motifs) mediate repression through PRC1 components such as AtBMI1 (ref. 15). Whereas teloboxes showed a quantitative correlation to the loss of H3K27me3 in trb1/2/3, RY motifs followed an opposite trend (Fig. 3f). Thus, the loss of H3K27me3 in trb1/2/3 was probably a direct effect of the loss of TRB binding at telobox-enriched target regions with comparably few RY motifs. In contrast, the gain of H3K27me3 in trb1/2/3 mutants may be explained by competition among different cis-encoded elements for PRC2 recruitment.

We tested whether FIE binding was dependent on the presence of TRBs at three previously described FIE- and TRB1-target genes17. Protoplasts of trb1/2/3 mutants transfected with FIE-GFP showed significantly less FIE binding than did Col-0 controls at SEP3, AG, and EMBRYONIC FLOWER 1 (EMF1), thus indicating that FIE binding was dependent on the presence of TRBs (Fig. 3g). In contrast, the presence of PRC2 and H3K27me3 was not a precondition for TRB1 binding, which was even at times increased in the clf-28 mutant compared to Col-0 (Fig. 3h and Supplementary Fig. 7).

Because our data suggested that TRBs directly recruit PRC2 to teloboxes, we tested the possibility of a physical interaction between TRBs and CLF or SWN. In vivo, TRB1-3 fused to GFP coimmunoprecipitated with hemagglutinin (HA)-tagged CLF or SWN in Arabidopsis protoplasts (Fig. 4a and Supplementary Fig. 8). Bimolecular fluorescence complementation (BiFC) in tobacco cells confirmed the interaction between TRB1 and CLF (Supplementary Fig. 8), whereas yeast two-hybrid (Y2H) assays detected an interaction of CLF or SWN with TRB2 and TRB3 but not TRB1 (Supplementary Fig. 9a,b). TRBs contain an N-terminal Myb domain, a central linker histone H1/H5 domain, and a C-terminal coiled-coil domain (Supplementary Fig. 9c). The Myb domain binds teloboxes, whereas the H1/H5 domain mediates interaction between TRBs29. The coiled-coil domain of TRB3 was found to interact with CLF or SWN, although neither the Myb nor the H1/H5 domains did (Supplementary Fig. 9d,e). As for TRB1, the coiled-coil domain of TRB1 did not interact with CLF or SWN (Supplementary Fig. 9f). Because TRBs form hetero- and/or homomultimers (Supplementary Fig. 9a and ref. 30), we suggest that the interaction between TRBs and CLF or SWN may involve TRB heteromultimers, particularly in the case of TRB1.

Fig. 4: Teloboxes recruit CLF–PRC2 via TRBs.
Fig. 4

a, Coimmunoprecipitation (IP) of GFP-TRB1 and HA-CLF with anti-GFP and anti-HA antibodies. A representative blot from three replicated experiments is shown (uncropped image in Supplementary Fig. 14). b, IGV screen shot of SEP3. c, Promoter reporter constructs used in transgenic lines carrying a wild-type or mutant (M) telobox. Black lines correspond to regions analyzed for TRB1-GFP binding (long line; Supplementary Fig. 12a) or H3K27me3 levels (short line; e). d, Quantification of GUS expression by qRT–PCR in wild-type or M transgenic lines. PP2A was used as an internal reference, and data were analyzed with the –ΔΔCt method. Plots show data for 4 independent lines. Central vertical lines and whiskers indicate mean ± s.d. Statistical analysis was performed with Mann–Whitney–Wilcoxon tests (df = 6). e, Quantification of H3K27me3 levels at SEP3 for GUS and endogenous SEP3, plotted as in d. Statistical analysis was performed with analysis of variance (ANOVA) and Holm–Sidak multiple-testing correction (df = 12). Different letters indicate distinct groups (two-tailed P values <0.05). f, Illustration of CaMV 35 S promoter–driven dCAS9-TRB1-HC constructs used for tethering truncated TRB1 to target regions. g, Western blot of HA-tagged dCAS9-TRB1-HC and dCAS9 protein coexpressed in protoplasts with sgRNAs against SEP3 (sgSEP3) and FT (sgFT). Anti-H3 was used as a loading control. A representative blot of six replicate transfection experiments is shown (uncropped image in Supplementary Fig. 14). h, Quantification of GUS expression via qRT–PCR in protoplasts transfected as in g. Data were analyzed and plotted as in d. Statistical significance was determined with two-tailed unpaired t tests (df = 10). i, Proposed model of PRC2 recruitment mediated by teloboxes and TRBs.

Teloboxes enhance the positive effect of adjacent site II motifs on transcription28; however, TRB- and telobox-dependent PRC2 recruitment should lead to transcriptional repression. We tested the effect of a telobox in the SEP3 promoter (SEP3pro) on transcription and chromatin regulation by using transgenic SEP3pro-GUS promoter reporter constructs (Fig. 4b,c). Independent lines of 10-d-old seedlings carrying SEP3pro-GUS with a wild-type telobox expressed almost no GUS in leaves, whereas plants carrying SEP3pro-GUS with a mutated telobox expressed GUS ectopically (Supplementary Fig. 10a–c). Both constructs were comparably expressed in the inflorescence (Supplementary Fig. 10d,e)31. We confirmed the semiquantitative histochemical GUS pattern in seedlings through qRT–PCR (Fig. 4d). Mutating the telobox resulted in a significant decrease in H3K27me3 levels compared with those in wild type at the corresponding transgenes (Fig. 4e).

Proximal to the telobox, the SEP3 promoter contains several telobox-related motifs, which can be bound by TRB1, according to ChIP–seq data, as well as by TRB2 and TRB3, according to ChIP–PCR data (Fig. 4b and Supplementary Fig. 11). We used transfected transgenic Arabidopsis protoplasts to test whether the telobox mutation at the SEP3 promoter decreased TRB1 binding. ChIP–PCR showed a lower enrichment of TRB1-GFP at the mutant SEP3pro-M-GUS than at the wild-type SEP3pro-WT-GUS in the corresponding transgenic protoplasts (Supplementary Fig. 12a). Likewise, TRB3 protein produced in bacteria showed low binding to a SEP3 promoter fragment containing the mutated telobox (Supplementary Fig. 12b). Together, our results indicated that all TRBs bind the telobox region at the SEP3 promoter (Supplementary Fig. 12 and refs 29,32,33). Mutation of the telobox element decreases rather than abolishes TRB binding, owing to the presence of redundant motifs, but this decrease is sufficient to impair repression of SEP3.

To determine whether SEP3 repression could be reconstituted by increasing the level of TRB1 at the mutagenized promoter, we tethered TRB1 to the mutated SEP3 promoter region. For this purpose, the H1/H5 and coiled-coil domains of TRB1 (TRB-HC) were fused to the C terminus of catalytically inactive CAS9 (dCAS9)34 (Fig. 4f) and coexpressed with a single guide RNA (sgRNA) recognizing the mutated telobox. Although dCAS9–TRB1-HC protein accumulated at much lower levels than dCAS9 protein in SEP3pro-M-GUS protoplasts (Fig. 4g), the GUS expression was lower than that in controls (Fig. 4h).

We propose that TRBs bind to the telobox and related motifs and recruit PRC2 for H3K27me3 deposition at target genes (Fig. 4i). For a subset of genes, TRBs are required to ensure stable H3K27me3 levels. Genes that are less dependent on the presence of TRBs may predominantly rely on different motifs, such as RY motifs, which recruit PRCs through an interaction with VAL family transcription factors11,12,15,16. A recent report has suggested that AZF1, in cooperation with BPC1 and GAGA motifs, recruits PRC2 to teloboxes17. Interestingly, the set of AZF1-target genes does not significantly overlap with TRB1-target or Col-0 biased genes, thus indicating that more than one pathway connects PRC2 to teloboxes (Supplementary Fig. 13). In conclusion, recruitment of Polycomb-group proteins in plants depends on several trans/cis-regulatory modules that may act in parallel or alone, in a manner dependent on the target genes’ repertoire of cis elements.


GEO Gene Expression Omnibus, https://www.ncbi.nlm.nih.gov/geo/; DDBJ Gene Bank Japan, https://trace.ddbj.nig.ac.jp/; WEBCAT WEB based chromatin association tester, https://www.biotools.fr/CAT/webCAT/; MA-norm, http://bioinfo.sibs.ac.cn/zhanglab/MAnorm/MAnorm.htm.


Plant materials and cultivation conditions

The trb1-2 (Salk_001540) and trb3-2 (Salk_134641) alleles were obtained from the SALK T-DNA-insertion-line collection (background accession Col-0)41. The trb2-1 (Flag_242F11) mutant in the Ws-2 background was obtained from the INRA T-DNA-insertion-line collection42. The ku70 and tert (G5) mutants were provided by K. Riha (Gregor Mendel Institute of Molecular Plant Biology). Oligonucleotide primers used for genotyping are indicated in Supplementary Table 1.

For qRT–PCR/RNA-seq and ChIP–qPCR/ChIP–seq, seeds of Col-0 and corresponding mutants were sterilized in 70% ethanol and sown on GM medium. Material was collected from 10-d-old seedlings grown in Percival growth cabinets at 22 °C (LD, 16 h light/8 h dark). For phenotypic analysis, seeds were sown on soil and transferred to LD conditions after stratification (4 °C, 3 d). Flowering time was determined in randomly distributed plants according to the number of rosette and cauline leaves of the main shoot, and plant size was measured as the largest rosette diameter at bolting time.

Plasmid construction, generation of transgenic plants, and histochemical GUS staining

For the TRB2pro-TRB2-Flag, TRB2pro-TRB2-YFP, and TRB3pro-TRB3-YFP constructs, 2-kb-upstream fragments and gene-body regions of TRB2 or TRB3 without stop codons were PCR-amplified from genomic DNA of Col-0 with GW-compatible primers (Supplementary Table 1). gTRB2 and gTRB3 were fused with a C-terminal Flag sequence in the pCAMBIA1305 vector or a YFP epitope sequence in the pXCG-mYFP vector. For the TRB2 CRISPR–CAS9 line, design and cloning of sgRNA was performed as previously described20. Briefly, oligonucleotide primers containing TRB2 sgRNA were used for sgRNA amplification, and the insert was cloned into pYB196 via the BamHI and SpeI sites (Supplementary Table 1). Transgenic plants were generated by Agrobacterium-mediated gene transfer with the floral dip method43.

For SEP3 expression, a 2.2-kb-long promoter sequence, located upstream of the ATG, was amplified from Col-0 genomic DNA and introduced into a GW::GUS-pGREEN vector to drive GUS reporter gene expression. Oligonucleotide primers used for cloning and telobox-motif mutation are listed in Supplementary Table 1. Transgenic plants were generated by Agrobacterium-mediated transfer with the floral dip method. 10-d-old seedlings or 35-d-old inflorescences of transgenic plants carrying SEP3pro-GUS with wild-type or mutated telobox elements were used for histochemical GUS staining, as previously described. The results of GUS staining were visualized under a light stereomicroscope (MZ 16 FA; Leica).

Terminal restriction fragment analysis

Terminal restriction fragment experiments were performed as previously described19. Briefly, 2 µg genomic DNA was extracted with a DNeasy Plant Mini Kit (Qiagen) from 10-d-old seedlings grown under LD conditions. The DNA was then digested by MseI (NEB) at 37 °C overnight. The digested DNA was electrophoresed on an agarose gel and blotted to a polyvinylidene fluoride membrane. Oligonucleotide (TTTAGGG) was end-labeled with T4 polynucleotide kinase and [γ-32P]ATP and used as a probe for Southern blotting.

RNA isolation, quantitative PCR and RNA-seq-library preparation

Total RNA was extracted with an RNeasy mini kit (Qiagen) according to the manufacturer’s instructions. Total RNA (5 μg) was treated with DNase I (DNA-free kit, Ambion). For qRT–PCR, cDNA was generated at 42 °C for 2 h with Superscript II reverse transcriptase and T18 oligonucleotide for priming (Life Technologies). Expression of TRB2 in the T-DNA line was determined through PCR with PP2A used as a control. qRT–PCR measurements were performed in a Bio-Rad iQ5 apparatus with iQ SYBR Green Supermix (Bio-Rad). Quantification was performed with the relative –ΔΔCt method, using PP2A for normalization. Oligonucleotide primers used for qRT–PCR are indicated in Supplementary Table 1.

For RNA-seq, material was collected from four independent biological replicates from Col-0 and trb1/2/3 mutants, respectively. DNA-free total RNA was generated as described above for the Illumina Tru-seq library preparation, which was carried out from DNA-free total RNA (3 μg) by the Max Planck Genome Centre in Cologne after oligo(dT)-based enrichment for mRNA.

ChIP and ChIP–seq library preparation

ChIP experiments were performed as previously described44 with anti-H3K27me3 (Millipore, 07-449) or anti-GFP (Abcam, ab290). Briefly, 10-d-old seedlings or Arabidopsis protoplasts were fixed in PBS buffer with 1% formaldehyde under vacuum two times for 10 min, after which the fixed seedlings were homogenized in liquid nitrogen. Chromatin was extracted and sonicated to produce DNA fragments of ~200–500 bp. H3K27me3- or YFP/GFP-associated DNA was enriched with protein A–Sepharose (GE) for antibody recovery. Primers used for ChIP–qPCR or ChIP–PCR are listed in Supplementary Table 1.

For ChIP–seq, two immunoprecipitations from independent biological replicates were processed for the NGS library preparation. All libraries were prepared with an Ovation Ultralow Library System (NuGEN) according to the manufacturer’s instructions, and 80% of a typical ChIP was used as starting material. After amplification for 16 PCR cycles, DNA between 200 and 300 bp length was purified from an agarose gel. An aliquot of the library was tested before and after PCR to confirm amplification by quantitative PCR. Sequencing was performed as single-end 100-nt reads on the Illumina HiSeq platform by the Max Planck Genome Centre in Cologne.

RNA-seq and ChIP–seq data analysis

RNA-seq data of Polycomb-group mutants (clf-29/swn-21, bmi1a/b, ring1a/b, clf-29, tfl2-2, and swn-21) were downloaded from Gene Expression Omnibus (GEO; see URLs) under accession number GSE67322 (ref. 23). siFIE microarray data were downloaded from the GEO under accession number GSE48857 (ref. 27). FIE and TRB1 ChIP–seq data were downloaded from the DNA Data Bank of Japan (see URLs) under accession numbers SRP027413 (ref. 27) and SRP058939 (ref. 26), respectively.

The raw sequencing reads were cleaned by removal of bases with low quality scores (<20) and irregular GC content, cutting sequencing adaptors and then filtering short reads. As a result, 12 million to 36 million reads with MAPQ >20 were obtained for further analyses. The cleaned reads were mapped to the Arabidopsis thaliana genome (TAIR10) with BWA 0.7.5a-r405 (ref. 45) for DNA sequencing and TOPHAT v2.0.9 (ref. 46) for RNA sequencing, both with default settings. SICER_V1.1 (ref. 47) was used to identify read-enriched regions for H3K27me3 ChIP–seq data without a background control. Next, MAnorm35 was applied to characterize the quantitative changes in H3K27me3 levels in trb1/2/3 compared with Col-0. MACS1.4 (ref. 48) was used for peak calling for TRB1 and FIE-HA ChIP–seq datasets using, respectively, a no-antibody ChIP control from Col-0 seedlings and an HA ChIP performed on nontransgenic C24 plants as a background. The target gene of each peak was defined as the gene closest to a given peak within 1 kb centered on the TSS. For IGV viewing, we normalized H3K27me3 samples such that the numbers of reads in peak regions in common between Col-0 and trb1/2/3 were the same.

Correlations between ChIP–seq datasets were tested with the Genome Association Tester (GAT)36 algorithm, as implemented in WebCAT (see URLs). GAT is a permutation approach in which overlaps between the fragments in each dataset are tested against 100 permutations of the same datasets across the genomic coordinates. Enrichment scores were calculated with the formula \(Z={{\rm{log}}}_{2}\left(\frac{{\rm{observed}}\,{\rm{overlap}}}{{\rm{expected}}\,{\rm{overlap}}}\right)\) (−log10(P value)) and are displayed as a heat map.

For quantitative comparison of RNA-seq datasets, the number of reads mapped to each gene was counted via HTseq-count49. Differentially expressed genes were detected with DESeq50, according to the combined criteria: |log2(fold change)| > 1 and adjusted P value < 0.05. To explore the relationships in transcriptomic changes between trb1/2/3 and Polycomb-group mutants, genes with differential expression in at least one of the 15 mutants (clf-29/swn-21, siFIE, trb1/2/3, bmi1a/b, ring1a/b, clf-29, tfl2-2, swn-21, lhp1 trb1/trb3, lhp1/trb1, lhp1/trb3, lhp1, trb3, trb1, and trb1/trb3) were collected, thus resulting in 6,289 genes, which were further clustered via hierarchical clustering. Fishers’ exact test51 was used to calculate the significance of enrichment between gene lists.

TF-binding-motif enrichment analysis

To detect TF-binding motifs enriched in H3K27me3 peak regions, we downloaded 924 position-weight matrixes from four major plant motif databases: JASPAR (228 motifs)38, AthaMap (183 motifs)40, AGRIS (91 motifs)37, and PLACE (422 motifs)39. We then performed a motif scan applied to a 1,000-bp window centered at the peak center. For each motif M, the raw motif matching score at each peak P was calculated as \(\mathop{\mathrm{max}}\limits_{S\subseteq P}\left[{\rm{log}}\frac{P\left(S{\rm{| }}M\right)}{P\left(S{\rm{| }}B\right)}\right]\), in which S is a sequence fragment of the same length as the motif, and B is the background frequency of the four nucleotides (A, C, G, and T), estimated from the genome. The enrichment of motif M in a peak list was defined as the ratio of the motif occurrence in the peak list as compared to its occurrence in random genomic regions. Fisher’s exact test was used to calculate the enrichment P value. Enriched motifs with an enrichment P value of 0.01 are presented in a heat map.

Transient expression and coimmunoprecipitation or RNA extraction

Full-length coding sequences for TRB1, TRB2, TRB3, CLF, and SWN were fused with C-terminal GFP (for TRB1–3) or N-terminal HA epitope (for CLF and SWN) sequences in the pAM-GW-GFP or pER8-HA-GW vectors for expression under control of the CaMV 35S promoter or β-estradiol-inducible promoter, respectively (Supplementary Table 1). The constructs were cotransformed into Arabidopsis mesophyll protoplasts through the polyethylene glycol method, as previously described52. For pER8-CLF or pER8-SWN, 5 μM β-estradiol (Sigma) was added to the W5 solution. For coimmunoprecipitation, the protoplasts were harvested 12 h after transformation and lysed in lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM KCl, 2 mM EDTA, 0.1% Triton X-100, 10% glycerol, and 5 mM DTT) with freshly added proteinase-inhibitor cocktail (Sigma, P9599). The lysate was centrifuged at 13,000 g at 4 °C for 10 min. The supernatant was incubated with GFP-trap beads (ChromoTek) for 2 h at 4 °C while rotating on a wheel. The beads were washed with lysis buffer five times, diluted in 2× SDS loading buffer, and boiled for 5 min before separation on SDS–PAGE. Immunoblots were probed according to standard procedures with anti-GFP (Abcam, ab290) and anti-HA (Abcam, ab9110). For RNA extraction, pYB196-35S-dCAS9-sgFT, pYB196-35S-dCAS9-sgSEP3 or pYB196-35S-dCAS9-TRB1-HC-sgSEP3 was cotransformed into Arabidopsis mesophyll protoplasts through the polyethylene glycol method. The protoplasts were harvested 12 h after transformation and lysed in buffer from the RNeasy Plant Mini Kit RNA (Qiagen).

Colocalization and BiFC assay

Coding sequences of TRB1, CLF and ATJ3 were PCR-amplified with GW-compatible specific primers (Supplementary Table 1) and recombined into the split YFP binary vectors RfA-sYFPn-pBatTL-B and RfA-sYFPcpBatTL-B, p113-sYFPc, or the expression vectors pAM-GW-GFP or CZN656-GW-RFP53. Agrobacterium tumefaciens strains carrying plasmids for BiFC and the p19 silencing suppressor were grown overnight at 28 °C in 10 mL selective YEP medium, collected by centrifugation, and resuspended in infiltration medium (1 mM MgCl2, 150 μg/mL acetosyringone, and 1 mM MES-KOH, pH 5.6). After incubation at 28 °C in darkness for 3 h, cells were infiltrated into the abaxial surface of 3-week-old Nicotiana benthamiana plants. The fluorescence signal of YFP, RFP, or GFP was observed and recorded with an LSM 700 confocal laser-scanning microscope (Carl Zeiss).

Yeast-two-hybrid analysis

Coding sequences of TRB1, TRB2, TRB3, CLF, and SWN or different domains of TRB1 and TRB3 were PCR-amplified with specific primers (Supplementary Table 1) and cloned into the Y2H vectors pGADT7 and pGBKT7 or pBridge (Clontech). Yeast competent cells (strain AH109, Clontech) were prepared with a Frozen-EZ Yeast Transformation II Kit (Zymo Research) according to the manufacturer’s instructions. Bait and prey plasmids, or blank plasmids pGADT7 and pGBKT7, or pBridge were cotransformed into yeast competent cells. SD –Leu –Trp and SD –Leu –Trp –His dropout media were used for selection.

Electrophoretic mobility shift assays

A full-length coding sequence for TRB3 was fused with an N-terminal histidine epitope in the pET28b vector, transformed into Escherichia coli BL21 cells, and induced with 1 mM IPTG to express at 16 °C overnight. His-TRB3 protein was purified with Ni–NTA agarose (Qiagen). A SEP3 promoter fragment with or without telobox mutation was amplified by PCR with Cy5–labeled primers. EMSA was performed with a LightShift Chemiluminescent EMSA Kit (Thermo Fisher Scientific) according to the manufacturer’s instructions. Briefly, Cy5-labeled PCR fragments were incubated with His-TRB3 protein at room temperature for 20 min, and free and bound DNA were separated in an acrylamide gel. Primers used for EMSA are listed in Supplementary Table 1.

Tethering experiments

Point mutations encoding the p.Asp10Ala and p.His840Ala34,54 amino acid changes were introduced to the CAS9 gene to create a dCAS9 sequence in the pYB196 vector20. The ICU2 promoter of the pYB196 vector was replaced with the CaMV 35S promoter to drive dCAS9 expression. Coding sequences of the H1/H5 and coiled-coil domains of TRB1 were PCR-amplified with specific primers (Supplementary Table 1) and inserted into the pYB196 vector between the C terminus of dCAS9 and the HA tag. Oligonucleotide primers containing SEP3 sgRNA or FT sgRNA were used for sgRNA amplification, and the inserts were cloned into pYB196 via the BamHI and SpeI sites. The constructs were cotransformed into Arabidopsis mesophyll protoplasts prepared from seedlings stably transformed with SEP3pro-M-GUS through the polyethylene glycol method as previously described52.

Statistical analysis

All analysis of statistical power was performed post hoc. For sample sizes n ≥ 5, tests for statistical significance were performed with one-way ANOVA and multiple-comparison correction through the Holm–Sidak method (P < 0.05), after equal variance (Bartlett’s test) and normal distribution (Shapiro–Wilk test) were confirmed. For sample sizes n < 5 or in the absence of normal distribution, significance was determined with the nonparametric Mann–Whitney–Wilcoxon test, except when more than two groups were being considered.

Code availability

The computer code used for MA analysis is available for download (see URLs).

Reporting Summary

Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

Data availability

All newly generated RNA-seq and ChIP–seq sequencing data have been deposited in the European Nucleotide Archive (ENA). H3K27me3 ChIP–seq and RNA-seq data for Col-0 and trb1/2/3 have been deposited under umbrella accession PRJEB19936, and RNA-seq data for Col-0, trb1, trb3, and trb1/3 in the Col-0 or lhp1-4 background have been deposited under umbrella accession PRJEB8944.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.


  1. 1.

    Lewis, E. B. A gene complex controlling segmentation in Drosophila. Nature 276, 565–570 (1978).

  2. 2.

    Mozgova, I. & Hennig, L. The polycomb group protein regulatory network. Annu. Rev. Plant Biol. 66, 269–296 (2015).

  3. 3.

    Xiao, J. & Wagner, D. Polycomb repression in the regulation of growth and development in Arabidopsis. Curr. Opin. Plant Biol. 23, 15–24 (2015).

  4. 4.

    Simon, J. A. & Kingston, R. E. Occupying chromatin: Polycomb mechanisms for getting to genomic targets, stopping transcriptional traffic, and staying put. Mol. Cell 49, 808–824 (2013).

  5. 5.

    Kassis, J. A. & Brown, J. L. Polycomb group response elements in Drosophila and vertebrates. Adv. Genet. 81, 83–118 (2013).

  6. 6.

    Förderer, A., Zhou, Y. & Turck, F. The age of multiplexity: recruitment and interactions of Polycomb complexes in plants. Curr. Opin. Plant Biol. 29, 169–178 (2016).

  7. 7.

    Derkacheva, M. et al. Arabidopsis MSI1 connects LHP1 to PRC2 complexes. EMBO J. 32, 2073–2085 (2013).

  8. 8.

    Liang, S. C. et al. Kicking against the PRCs: a domesticated transposase antagonises silencing mediated by Polycomb group proteins and is an accessory component of Polycomb repressive complex 2. PLoS Genet. 11, e1005660 (2015).

  9. 9.

    Molitor, A. M., Bu, Z., Yu, Y. & Shen, W. H. Arabidopsis AL PHD-PRC1 complexes promote seed germination through H3K4me3-to-H3K27me3 chromatin state switch in repression of seed developmental genes. PLoS Genet. 10, e1004091 (2014).

  10. 10.

    Lanzuolo, C. & Orlando, V. Memories from the polycomb group proteins. Annu. Rev. Genet. 46, 561–589 (2012).

  11. 11.

    Qüesta, J. I., Song, J., Geraldo, N., An, H. & Dean, C. Arabidopsis transcriptional repressor VAL1 triggers Polycomb silencing at FLC during vernalization. Science 353, 485–488 (2016).

  12. 12.

    Yuan, W. et al. A cis cold memory element and a trans epigenome reader mediate Polycomb silencing of FLC by vernalization in Arabidopsis. Nat. Genet. 48, 1527–1534 (2016).

  13. 13.

    Berger, N., Dubreucq, B., Roudier, F., Dubos, C. & Lepiniec, L. Transcriptional regulation of Arabidopsis. LEAFY COTYLEDON2 involves RLE, a cis-element that regulates trimethylation of histone H3 at lysine-27. Plant Cell 23, 4065–4078 (2011).

  14. 14.

    Lodha, M., Marco, C. F. & Timmermans, M. C. P. The ASYMMETRIC LEAVES complex maintains repression of KNOX homeobox genes via direct recruitment of Polycomb-repressivecomplex2. Genes Dev. 27, 596–601 (2013).

  15. 15.

    Merini, W. et al. The Arabidopsis Polycomb repressive complex 1 (PRC1) components AtBMI1A, B, and C impact gene networks throughout all stages of plant development. Plant Physiol. 173, 627–641 (2017).

  16. 16.

    Yang, C. et al. VAL- and AtBMI1-mediated H2Aub initiate the switch from embryonic to postgerminative growth in Arabidopsis. Curr. Biol. 23, 1324–1329 (2013).

  17. 17.

    Xiao, J. et al. Cis and trans determinants of epigenetic silencing by Polycomb repressive complex 2 in Arabidopsis. Nat. Genet. 49, 1546–1552 (2017).

  18. 18.

    Hecker, A. et al. The Arabidopsis GAGA-binding factor BASIC PENTACYSTEINE6 recruits the POLYCOMB-REPRESSIVE COMPLEX1 component LIKE HETEROCHROMATIN PROTEIN1 to GAGA DNA motifs. Plant Physiol. 168, 1013–1024 (2015).

  19. 19.

    Zhou, Y., Hartwig, B., James, G. V., Schneeberger, K. & Turck, F. Complementary activities of TELOMERE REPEAT BINDING proteins and Polycomb group complexes in transcriptional regulation of target genes. Plant Cell 28, 87–101 (2016).

  20. 20.

    Hyun, Y. et al. Site-directed mutagenesis in Arabidopsis thaliana using dividing tissue-targeted RGEN of the CRISPR/Cas system to generate heritable null alleles. Planta 241, 271–284 (2015).

  21. 21.

    Schrumpfová, P. P. et al. Telomere repeat binding proteins are functional components of Arabidopsis telomeres and interact with telomerase. Plant J. 77, 770–781 (2014).

  22. 22.

    Lee, W. K. & Cho, M. H. Telomere-binding protein regulates the chromosome ends through the interaction with histone deacetylases in Arabidopsis thaliana. Nucleic Acids Res. 44, 4610–4624 (2016).

  23. 23.

    Wang, H. et al. Arabidopsis flower and embryo developmental genes are repressed in seedlings by different combinations of Polycomb group proteins in association with distinct sets of cis-regulatory elements. PLoS Genet. 12, e1005771 (2016).

  24. 24.

    Turck, F. et al. Arabidopsis TFL2/LHP1 specifically associates with genes marked by trimethylation of histone H3 lysine 27. PLoS Genet. 3, e86 (2007).

  25. 25.

    Chanvivattana, Y. et al. Interaction of Polycomb-group proteins controlling flowering inArabidopsis. Development 131, 5263–5276 (2004).

  26. 26.

    Schrumpfová, P. P. et al. Telomere binding protein TRB1 is associated with promoters of translation machinery genes in vivo. Plant Mol. Biol. 90, 189–206 (2016).

  27. 27.

    Deng, W. et al. Arabidopsis Polycomb Repressive Complex 2 binding sites contain putative GAGA factor binding motifs within coding regions of genes. BMC Genomics 14, 593 (2013).

  28. 28.

    Trémousaygue, D. et al. Internal telomeric repeats and ‘TCP domain’ protein-binding sites co-operate to regulate gene expression in Arabidopsis thaliana cycling cells. Plant J. 33, 957–966 (2003).

  29. 29.

    Mozgová, I., Schrumpfová, P. P., Hofr, C. & Fajkus, J. Functional characterization of domains in AtTRB1, a putative telomere-binding protein in Arabidopsis thaliana. Phytochemistry 69, 1814–1819 (2008).

  30. 30.

    Kuchar, M. & Fajkus, J. Interactions of putative telomere-binding proteins in Arabidopsis thaliana: identification of functional TRF2 homolog in plants. FEBS Lett. 578, 311–315 (2004).

  31. 31.

    Liu, C., Xi, W., Shen, L., Tan, C. & Yu, H. Regulation of floral patterning by flowering time genes. Dev. Cell 16, 711–722 (2009).

  32. 32.

    Lee, W. K., Yun, J. H., Lee, W. & Cho, M. H. DNA-binding domain of AtTRB2 reveals unique features of a single Myb histone protein family that binds to both Arabidopsis- and human-type telomericDNA sequences. Mol. Plant 5, 1406–1408 (2012).

  33. 33.

    Yun, J. H. et al. Solution structure of telomere binding domain of AtTRB2 derived from Arabidopsis thaliana. Biochem. Biophys. Res. Commun. 452, 436–442 (2014).

  34. 34.

    Sander, J. D. & Joung, J. K. CRISPR-Cas systems for editing, regulating and targeting genomes. Nat. Biotechnol. 32, 347–355 (2014).

  35. 35.

    Shao, Z., Zhang, Y., Yuan, G. C., Orkin, S. H. & Waxman, D. J. MAnorm: a robust model for quantitative comparison of ChIP-Seq data sets. Genome Biol. 13, R16 (2012).

  36. 36.

    Heger, A., Webber, C., Goodson, M., Ponting, C. P. & Lunter, G. GAT: a simulation framework for testing the association of genomic intervals. Bioinformatics 29, 2046–2048 (2013).

  37. 37.

    Yilmaz, A. et al. AGRIS: the Arabidopsis Gene Regulatory Information Server, an update. Nucleic Acids Res. 39, D1118–D1122 (2011).

  38. 38.

    Mathelier, A. et al. JASPAR 2014: an extensively expanded and updated open-access database of transcription factor binding profiles. Nucleic Acids Res. 42, D142–D147 (2014).

  39. 39.

    Higo, K., Ugawa, Y., Iwamoto, M. & Korenaga, T. Plant cis-acting regulatory DNA elements (PLACE) database: 1999. Nucleic Acids Res. 27, 297–300 (1999).

  40. 40.

    Hehl, R. & Bülow, L. AthaMap web tools for the analysis of transcriptional and posttranscriptional regulation of gene expression in Arabidopsis thaliana. Methods Mol. Biol. 1158, 139–156 (2014).

  41. 41.

    Alonso, J. M. et al. Genome-wide insertional mutagenesis of Arabidopsis thaliana. Science 301, 653–657 (2003).

  42. 42.

    Samson, F. et al. FLAGdb/FST: a database of mapped flanking insertion sites (FSTs) of Arabidopsis thaliana T-DNA transformants. Nucleic Acids Res. 30, 94–97 (2002).

  43. 43.

    Clough, S. J. & Bent, A. F. Floral dip: a simplified method for Agrobacterium-mediated transformation of Arabidopsis thaliana. Plant J. 16, 735–743 (1998).

  44. 44.

    Reimer, J. J. & Turck, F. Genome-wide mapping of protein-DNA interaction by chromatin immunoprecipitation and DNA microarray hybridization (ChIP-chip). Part A: ChIP-chip molecular methods. Methods Mol. Biol. 631, 139–160 (2010).

  45. 45.

    Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009).

  46. 46.

    Kim, D. et al. TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions. Genome Biol. 14, R36 (2013).

  47. 47.

    Zang, C. et al. A clustering approach for identification of enriched domains from histone modification ChIP-Seq data. Bioinformatics 25, 1952–1958 (2009).

  48. 48.

    Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, R137 (2008).

  49. 49.

    Anders, S., Pyl, P. T. & Huber, W. HTSeq: a Python framework to work with high-throughput sequencing data. Bioinformatics 31, 166–169 (2015).

  50. 50.

    Anders, S. & Huber, W. Differential expression analysis for sequence count data. Genome Biol. 11, R106 (2010).

  51. 51.

    Fisher, R. A. A new test for 2 × 2 tables. Nature 156, 388 (1945).

  52. 52.

    Sheen, J. Signal transduction in maize and Arabidopsis mesophyll protoplasts. Plant Physiol. 127, 1466–1475 (2001).

  53. 53.

    Hu, C. D. & Kerppola, T. K. Simultaneous visualization of multiple protein interactions in living cells using multicolor fluorescence complementation analysis. Nat. Biotechnol. 21, 539–545 (2003).

  54. 54.

    Ran, F. A. et al. Double nicking by RNA-guided CRISPR Cas9 for enhanced genome editing specificity. Cell 154, 1380–1389 (2013).

Download references


We thank P. Taenzler for excellent technical help and A. Brazel for critical reading of the manuscript. This work was supported by core funding from the Max Planck Society (Y. Zhou, K.K., J.A.D., and F.T.), the Chinese Scholarship Council (T.Y.) and the National Natural Science Foundation of China (grant no. 31570319; Y.W. and Y. Zhang). We thank K. Riha (Gregor Mendel Institute of Molecular Plant Biology) and J. Goodrich (University of Edinburgh) for providing materials.

Author information


  1. Max Planck Institute for Plant Breeding Research, Department of Plant Developmental Biology, Köln, Germany

    • Yue Zhou
    • , Kristin Krause
    • , Tingting Yang
    •  & Franziska Turck
  2. National Key Laboratory of Plant Molecular Genetics, CAS Center for Excellence in Molecular Plant Sciences, Institute of Plant Physiology and Ecology, Shanghai Institute for Biological Sciences, Shanghai, China

    • Yuejun Wang
    •  & Yijing Zhang
  3. University of Chinese Academy of Sciences, Shanghai, China

    • Yuejun Wang
    •  & Yijing Zhang
  4. Max Planck Institute for Plant Breeding Research, Department of Plant Microbe Interactions, Köln, Germany

    • Joram A. Dongus


  1. Search for Yue Zhou in:

  2. Search for Yuejun Wang in:

  3. Search for Kristin Krause in:

  4. Search for Tingting Yang in:

  5. Search for Joram A. Dongus in:

  6. Search for Yijing Zhang in:

  7. Search for Franziska Turck in:


Y. Zhou performed and supervised all experimental work; K.K. performed the functional analysis of the SEP3 gene; J.A.D. assisted in Y2H assays and coimmunoprecipitation; K.K. and J.A.D. contributed to the development of theTRB2 CRISPR–CAS9 line; K.K. and T.Y. contributed the protoplast work; Y. Zhang and Y.W. carried out all bioinformatics analysis; and Y. Zhou, K.K., and F.T. planned the study and wrote the manuscript.

Competing interests

The authors declare no competing interests.

Corresponding author

Correspondence to Franziska Turck.

Integrated supplementary information

  1. Supplementary Figure 1 Phenotypes of trb, clf, and trb clf mutants.

    (a) Phenotype of 28-days-old Col-0, trb1-2, trb3-2, trb1-2 trb3-2, clf-28, trb1-2 clf-28, trb3-2 clf-28 and trb1-2 trb3-2 clf-2 plants grown at 22 °C in long-day (LD) conditions. Scale bar indicates 1 cm. (b) Flowering time scored for genotypes in (a) as number of leaves and average of rosette size at bolting. Error bars indicate mean ± s.e.m. (n = 9). Statistical significance was determined by one-way analysis of variance (ANOVA) and multiple comparison by the Holm–Sidak method (P < 0.05). Different letters indicate significance groups. (c) Phenotype of 28-day-old Ws-2, trb2-1, clf-50 and trb2-1 clf-50 plants grown at 22 °C in LD conditions. Scale bar indicates 1 cm. (d and e) Flowering time scored for genotypes in (c) as number of leaves and average rosette size at bolting. Significance tested as in (b). (f) Expression of TRB2 in Ws-2 and trb2-1 mutant backgrounds measured by qRT-PCR. PP2A was used as reference gene. The presence of full-length transcripts was analyzed.

  2. Supplementary Figure 2 Phenotype of higher-order trb mutants and complementation of triple trb1/2/3.

    (a) Phenotype of 10-day-old seedlings of Col-0, trb1/2, trb1/3 and trb2/3 mutants grown at 22 °C in LD conditions. Scale bar: 1 cm. (b) Phenotype of 10-day-old seedlings of Col-0, trb1/2/3, and gTRB2-Flag transgenic lines (L12 and L21) in trb1/2/3 background. Scale bar: 1 cm. (c) +1 insertion (2nd exon of TRB2) in trb2-2 line is identified in trb1/3 double mutant background. The predicted DNA cleavage site is marked with an arrow. (d) Phenotype of 10-day-old seedlings of Col-0, trb1/2/3 and trb2-2 trb1/3 mutants. Scale bar: 1 cm.

  3. Supplementary Figure 3 Nuclear localization of TRB proteins and their effect on teleomere integrity.

    (a) Co-localization of TRB2 and TRB1 in the nucleus via transient expression of TRB2-GFP and TRB1-RFP in Tobacco leaves. Scale bar indicates 10 μM. (b) Analysis of pixel density along the axis indicated by a white line in (a) was performed by ImageJ. GFP signal intensities are shown in green and RFP signals in red. (c,d) gTRB2-YFP (c) and gTRB3-YFP (d) fusion proteins are detected in the nuclei of almost all cells in transgenic Arabidopsis plants. Yellow fluorescence detected in the root tip (upper panel) and the first true leaves (lower panel). Scale bars: 50 µm. (e) Telomere length measured by TRF analysis of genomic DNA prepared from 10-day-old Col-0, Ws-2, ku70, fifth generation (G5) tert, second (G2), and third (G3) generation of trb1/2/3 or trb1/3 trb2-/+ seedlings with ku70 and tert being included as reference.

  4. Supplementary Figure 4 Correlation between H3K27me3 and expression changes for genes misregulated in trb1/2/3.

    (a) Boxplot shows H3K27me3 ChIP-seq read intensity across genes commonly down- or up-regulated in trb123 and clf swn compared to all annotated genes. Venn diagrams are a reproduced from Fig. 1d for better clarity. A Hypergeometric test was used for statistical analysis. (b) Scatter plot showing the negative correlation between H3K27me3 and expression changes for genes mis-regulated in trb1/2/3.

  5. Supplementary Figure 5 Quantitative PCR validation of RNA-seq and H3K27me3 ChIP–seq data in Col-0 and trb1/2/3 seedlings.

    (a) Genes up-regulated in trb1/2/3 mutants. (b) Genes down-regulated in trb1/2/3 mutants. RNA was extracted from 10 day-old seedlings grown at 22 °C in LDs. Values were normalized to the reference gene PP2A using the –ΔΔCT approach. Vertical central line and whiskers indicate mean ± s. d. of four independently grown and processed replicate experiments used for qRT-PCR, symbols indicate data points as averages of three technical qPCR replicates. Statistical significance was determined by a Mann–Whitney–Wilcoxon test (df = 6) (c) H3K27me3 level change in Col-0 biased genes. (d) H3K27me3 level change in trb1/2/3 biased genes. (e) H3K27me3 level in common genes. Vertical central line and whiskers indicate mean ± s. d. of three independently grown and processed replicate experiments used for ChIP-qPCR, symbols represent averages of three technical qPCR replicates. Statistical significance was determined by a Mann–Whitney–Wilcoxon test (df = 4).

  6. Supplementary Figure 6 List of cis elements enriched in different H3K27me3 gene categories detected in trb1/2/3 mutants.

    Sequence logos of the 15 most enriched elements detected by MEME analysis.

  7. Supplementary Figure 7 ChIP–qPCR analysis of H3K27me3 coverage in transiently FIE-GFP- and TRB1-GFP-expressing protoplasts.

    (a) Western blot of GFP tagged FIE or TRB1 proteins expressed in protoplasts prepared from Col-0 and trb1/2/3-2 or Col-0 and clf-28. Anti-H3 was used as a loading control. Displayed is a representative blot of three replicated experiments. See Supplementary Fig. 14 for full scan of cropped image. (b) H3K27me3 level at SEP3, EMF1, AG and ACT2 in protoplast prepared from Col-0 and clf-28 seedlings. ACT2 was used as negative control for H3K27me3 distribution. Vertical central line and whiskers indicate mean ± s. d. of ChIP-qPCR of three independently transfected and processed batches of protoplasts. Statistical significance was determined by Mann–Whitney–Wilcoxon test (df = 4).

  8. Supplementary Figure 8 Interaction of TRBs with CLF and SWN.

    (a) TRB1-nYFP co-infiltrated with cYFP-CLF (1) or ATJ3 (2) and ATJ3 co-infiltrated with cYFP-CLF (3) in Nicotiana tabacum leaves. LHP1-RFP was used as reference for co-infiltration and ATJ3 was used as control for interactions. Scale bar = 10 μM. (b) Percentage of RFP expressing nuclei as in (a) expressing YFP and RFP signal. Total cells = 100, counted in 5 independently infiltrated leaves; Statistical analysis was a one-tailed Mann–Whitney U-test (df = 8). Vertical central line and whiskers indicate mean ± s. d. (c) Co-immunoprecipitation of CLF and (b) SWN with TRB1-3 expressed in Arabidopsis mesophyll protoplasts. TRB1-GFP, TRB2-GFP or TRB3-GFP were immunoprecipitated with anti-GFP trap beads from protoplasts co-transfected with either HA-CLF or HA-SWN. The precipitates were analyzed by western blotting with anti-GFP and anti-HA antibodies. See Supplementary Fig. 14 for full scan of cropped images.

  9. Supplementary Figure 9 Interaction of TRBs with CLF and SWN in yeast cells.

    (a) GAL4 DNA-binding domain (BD) or BD-CLF and BD-SWN fusion proteins were tested for their ability to bind to an activation domain (AD) fused to different TRB proteins. (b) Mutual interactions between TRB1-3 were tested as in (a). (c) Schematic presentation of the TRB protein domain structure. (d-e) Interaction between the MYB, H1/5, or Coiled-coil domain of TRB3 and CLF (d) and SWN (e). (f) Interaction between the MYB, H1/5, or Coiled-coil domain of TRB1 with SWN. Growth shown for three dilutions (10−1, 10−2 and 10−3) of the yeast culture on SD medium lacking LW (-LW) and LWH (-LWH) in all panels.

  10. Supplementary Figure 10 Histochemical detection of GUS activity in transgenic SEP3pro-GUS lines with or without an intact telobox.

    (a) 10-day-old seedlings carrying wild type pSEP3pro-WT-GUS. (b) 10-day-old seedlings carrying mutated SEP3pro-M-GUS. Picture in (a-b) show representative pattern observed in four single copy insertion lines transformed in the Col-0 background per construct and (c) an untransformed Col-0 control. (d) GUS activity at the inflorescence of representative SEP3pro-WT-GUS and (e) SEP3pro-M-GUS line. Scale bars in all panels indicate 1 mm.

  11. Supplementary Figure 11 ChIP–qPCR analysis of TRB2 or TRB3 binding to the SEP3 promoter.

    Distal promoter region of FT was used to as negative region for TRB2 or TRB3 binding. Bars display the mean of three biological replicates used for ChIP-qPCR, Vertical central line and whiskers indicate mean ± s. d. Statistical significance was determined by Mann–Whitney–Wilcoxon test (df = 4).

  12. Supplementary Figure 12 TRB binding is decreased at a SEP3 promoter carrying a mutated telobox element.

    (a) ChIP-PCR of TRB1-GFP binding to pSEP3pro-WT-GUS or pSEP3pro-M-GUS transgenic lines; empty plasmids were used as a negative transfection control. Three independent biological experiments are shown. Note that PCR fragment necessary to distinguish the endogenous and transgenic SEP3 promoter is 800 bp long and therefore not compatible with qPCR. (b) Coomassie brilliant blue staining of recombinant His-TRB3 protein purified from E.coli. (c) Illustration depicting the SEP3 promoter region. Blue line indicates telobox; yellow lines indicate four telobox-related motifs; black line indicates fragment used for EMSA. (d) TRB3 binding to Cy5 labelled PCR fragments of SEP3 promoter with WT or mutated telobox.

  13. Supplementary Figure 13 AZF1- and TRB1-target genes do not show significant overlap.

    Numbers of significant AZF1 and TRB1 target genes overlapping with Col-0 biased genes. The p-value of the only significant overlap is indicated (Hypergeometric test).

  14. Supplementary Figure 14

    Original uncropped scans of representative western blots displayed in this manuscript.

Supplementary information

  1. Supplementary Text and Figures

    Supplementary Figures 1–14 and Supplementary Table 1

  2. Reporting Summary

  3. Supplementary Data Set 1: Summary table of gene expression and H3K27me3 changes in trb1/2/3 for all Arabidopsis genes annotated in TAIR10

    Sheet 1: Columns indicate 1: Presence or absence of teloboxes, 2: differential expression in trb1/2/3, 3: differential expression in clf/swn, 4: TRB1 binding according to Zhou et al. (2016)5, 5: TRB1 binding according to the present analysis, 6: H3K27me3 change in trb1/2/3 seedlings. Sheet 2: Gene expression fold-changes in trb1/2/3, clf-29swn-21 and siFIE mutants.

  4. Supplementary Data Set 2: Summary table for GO enrichment analysis of genes differentially expressed in trb1/2/3 seedlings

    Columns indicate 1: Cluster type, 2: enriched GO-term, 3: GO-term description, 4: GO-term count, 5: %, 6: p-value, 7: AGI codes, 8: length of list, 9: hits in list, 10: total gene list, 11: enrichment, 12: Bonferroni corrected p-value, 13: Benjamini corrected p-value, 14: FDR corrected p-value.

  5. Supplementary Data Set 3: Summary table of MAnorm analysis for regions differentially enriched in H3K27me3

    Columns indicate 1: Chromosome, 2: start of peak, 3: end of peak, 4: position of peak summit, 5: M-value, 6: A-value, 7: pvalue, 8: peak category, 9: normalized reads, 10: normalized read density in trb1/2/3.

  6. Supplementary Data Set 4: Summary table for GO enrichment analysis of Col-0 biased and trb1/2/3 biased genes

    Columns indicate 1: Cluster type, 2: enriched GO-term, 3: GO-term description, 4: GO-term count, 5: %, 6: p-value, 7: AGI codes, 8: length of list, 9: hits in list, 10: total gene list, 11: enrichment, 12: Bonferroni corrected p-value, 13: Benjamini corrected p-value, 14: FDR corrected p-value.

  7. Supplementary Data Set 5: Genomic coordinates of FIE and TRB1 enriched peaks

    Columns indicate 1: Chromosome, 2: start of peak, 3: end of peak, 4: length, 5: position of peak summit, 6: read tags, 7: corrected p-Value, 8: enrichment.

About this article

Publication history