Complementary Activities of TELOMERE REPEAT BINDING Proteins and Polycomb Group Complexes in Transcriptional Regulation of Target Genes

TELOMERE REPEAT BINDING proteins sustain repression of target genes they share with Polycomb Group repressive complexes.

Abstract

In multicellular organisms, Polycomb Repressive Complex 1 (PRC1) and PRC2 repress target genes through histone modification and chromatin compaction. Arabidopsis thaliana mutants strongly compromised in the pathway cannot develop differentiated organs. LIKE HETEROCHROMATIN PROTEIN1 (LHP1) is so far the only known plant PRC1 component that directly binds to H3K27me3, the histone modification set by PRC2, and also associates genome-wide with trimethylation of lysine 27 of histone H3 (H3K27me3). Surprisingly, lhp1 mutants show relatively mild phenotypic alterations. To explain this paradox, we screened for genetic enhancers of lhp1 mutants to identify novel components repressing target genes together with, or in parallel to, LHP1. Two enhancing mutations were mapped to TELOMERE REPEAT BINDING PROTEIN1 (TRB1) and its paralog TRB3. We show that TRB1 binds to thousands of genomic sites containing telobox or related cis-elements with a significant increase of sites and strength of binding in the lhp1 background. Furthermore, in combination with lhp1, but not alone, trb1 mutants show increased transcription of LHP1 targets, such as floral meristem identity genes, which are more likely to be bound by TRB1 in the lhp1 background. By contrast, expression of a subset of LHP1-independent TRB1 target genes, many involved in primary metabolism, is decreased in the absence of TRB1 alone. Thus, TRB1 is a bivalent transcriptional modulator that maintains downregulation of Polycomb Group (PcG) target genes in lhp1 mutants, while it sustains high expression of targets that are regulated independently of PcG.

INTRODUCTION

Polycomb Group (PcG) proteins epigenetically regulate cell fate and identity in higher eukaryotes by chromatin-mediated gene repression. PcG proteins form functionally distinct complexes that act in concert to modify chromatin by trimethylation of lysine 27 of histone H3 (H3K27me3) and monoubiquitination of a lysine residue within the histone fold domain of H2A (H2Aub). Occurrence of these two hallmark modifications leads to local chromatin compaction and gene repression by mechanisms that are not yet entirely understood (Schuettengruber and Cavalli, 2009; Margueron and Reinberg, 2011; Xiao and Wagner, 2015).

In Arabidopsis thaliana, two Polycomb repressive complexes (PRCs) have been identified, both of which exist in multiple variants due to gene family expansion and functional diversification (Derkacheva and Hennig, 2014). PRC2 sets the H3K27me3 mark, which covers around 4400 to 7000 genes, depending on tissue type and target threshold definition used in several independent genome-wide studies (Zhang et al., 2007b; Weinhofer et al., 2010; Lafos et al., 2011; Dong et al., 2012). PRC1 fulfils three molecular functions that may not depend on a single holocomplex, but could be implemented by PRC1-like subcomplexes (Mozgova and Hennig, 2015). PRC1 component LIKE HETEROCHROMATIN PROTEIN1 (LHP1) can recognize H3K27me3 through its chromodomain (Turck et al., 2007; Zhang et al., 2007a; Exner et al., 2009), while H2A monoubiquitination is dependent on the presence of RING-RAWUL twin domain proteins of the Arabidopsis B Lymphoma Mo-MLV Insertion Region 1 and RING FINGER PROTEIN1 subfamily (Bratzel et al., 2010; Yang et al., 2013; Calonje, 2014). Finally, EMBRYONIC FLOWER1 (EMF1) is a PRC1 component likely involved in chromatin compaction (Calonje et al., 2008; Beh et al., 2012). Genome-wide binding studies have been performed for both LHP1 (Turck et al., 2007; Zhang et al., 2007a; Engelhorn et al., 2012) and EMF1 (Kim et al., 2012) and suggested an overlapping binding pattern strictly correlated to the occurrence of H3K27me3.

LHP1 loss-of-function mutants show pleiotropic phenotypes, including early flowering, upward leaf curling, reduced leaf size, and dwarfism (Larsson et al., 1998; Gaudin et al., 2001; Kotake et al., 2003). The collective upregulation of MADS domain transcription factors encoded by the ABCDE floral meristem identity genes, such as SEPALLATA3 (SEP3), APETALA1 (AP1), AP3, AGAMOUS (AG) and SHATTERPROOF1, explains part of the lhp1 mutant phenotype, although it is rather difficult to disentangle the effects directly caused by the lack of LHP1-mediated repression from those due to the mutual upregulation within the regulatory network (Nakahigashi et al., 2005; Derkacheva et al., 2013). In addition, both the floral repressor FLOWERING LOCUS C (FLC) and FLOWERING LOCUS T (FT), encoding the photoperiod-dependent florigen, are upregulated in lhp1 mutants (Takada and Goto, 2003; Mylne et al., 2006; Sung et al., 2006). FLC is a direct repressor of FT, but in the lhp1 mutant, FT is upregulated in phloem companion cells despite increased FLC levels, explaining most of the early flowering phenotype (Kotake et al., 2003; Searle et al., 2006; Farrona et al., 2011b).

LHP1 directly interacts with RING-RAWUL proteins and EMF1, indicating that LHP1 can be present in several PRC1-like complexes (Xu and Shen, 2008; Bratzel et al., 2010). In addition, LHP1 was detected in pull-down experiments performed using epitope-tagged MSI1 as bait due to a direct interaction with MSI1 (Derkacheva et al., 2013). It was suggested that recruitment of PRC2 by H3K27me3-bound LHP1 is important to maintain H3K27me3 levels in root cultures undergoing rapid cell division (Derkacheva et al., 2013).

Recent reports suggest that PRC1 can at least sometimes act upstream of PRC2 since H2Aub preceded H3K27me3 during postgerminative repression of seed maturation genes (Bratzel et al., 2010; Yang et al., 2013; Calonje, 2014). In theory, LHP1 would be able to mediate the connection between PRC1-like complexes and PRC2 in the absence of H3K27me3 in this scenario. Moreover, in vitro, Arabidopsis GAGA-motif binding factors, such as BASIC PENTACYSTEINE6 (BPC6), can recruit PRC1 to GAGA motifs by their direct interaction with LHP1, which may subsequently recruit PRC2 (Hecker et al., 2015). Other transcription factors, such as SHORT VEGETATIVE PHASE and SHORT ROOT, were shown to interact with LHP1 and, likewise, may be involved in triggering PcG-mediated repression at their target genes (Cui and Benfey, 2009; Liu et al., 2009).

Despite the central position of LHP1 in PRC1 and its high connectivity to PRC2, the role of LHP1 within the PcG pathway remains a conundrum. LHP1, in contrast to most PRC1 and 2 components, is encoded by a single copy gene in Arabidopsis, for which true null alleles are available. Nevertheless, the lhp1 phenotype is mild compared with other more severe PcG mutants (Mozgova and Hennig, 2015). For example, a combination of mutations in CURLY LEAF and SWINGER, which together provide all sporophytic H3K27me3-directed activity (Farrona et al., 2011b; Lafos et al., 2011), leads to much more severe developmental defects. After germination, clf swn mutant plants develop a transdifferentiating cell clump that initiates organ development, including the formation of somatic embryos, without ever progressing to organ maturity (PcG callus) (Chanvivattana et al., 2004). A PcG callus is also observed in Arabidopsis mutants severely affected in H2Aub activity, such as triple bmi1a bmi1b bmi1c or double ring1a ring1b mutants (Chen et al., 2010; Yang et al., 2013). These observations indicate that there may be other proteins or pathways acting either redundantly or in parallel with LHP1.

Here, we report on results of a forward genetic screen for genetic enhancers of the lhp1 phenotype that we performed to uncover such novel components. We mapped mutations in genes coding for Myb-family transcription factor TELOMERE REPEAT BINDING PROTEIN1 (TRB1) and the related TRB3 as causal enhancers of the lhp1 phenotype. We show by genome-wide expression (RNA-seq) and chromatin immunoprecipitation sequencing (ChIP-seq) analysis that TRB1, by binding to teloboxes and telobox-related cis-elements, plays a role in transcription regulation that is independent of its previously identified role in telomere maintenance (Schrumpfova et al., 2014). Interestingly, TRB1 assists LHP1 in the repression of common target genes, while for a set of genes predominantly involved in primary metabolism, TRB1 binding seems to be required to sustain high expression levels. Finally, we discuss several functional models that would fit our observations.

RESULTS

Identification of Genetic Enhancers of lhp1

We induced mutations in the lhp1-3 mutant background (from now lhp1) by ethyl methanesulfonate (EMS) and screened for genetic enhancers. Two nonallelic lhp1 enhancers showing earlier flowering and stronger leaf size reduction (Figures 1A to 1C) were mapped by fast isogenic mapping (Hartwig et al., 2012) to the TRB1 and TRB3 loci (Supplemental Table 1). Mapping of causative mutations was validated by full complementation of the enhancer mutants by the corresponding genomic fragments and a reproduction of the enhanced phenotype after introducing independent T-DNA insertion alleles trb1-2 and trb3-2 in the lhp1 background (Supplemental Figures 1A to 1E). Note that both T-DNA alleles express a partial transcript encoding the MYB domain of TRB1 and 3; therefore, partial activity of the encoded proteins may be retained (Supplemental Figure 1D).

Figure 1.

trb1 and trb3 Alleles Enhance the lhp1 Mutant Phenotype Independently of Their Role in Telomere Maintenance.

(A) Phenotype of Col-0, trb1-1, trb3-1, trb1-1 trb3-1, lhp1-3, trb1-1 lhp1-3, trb3-1 lhp1-3, and trb1-1 trb3-1 lhp1-3 plants 28 d postgermination. Plants were grown at 22°C in long days. Bar = 1 cm.

(B) Flowering time of genotypes grown as in (A) scored as number of leaves. Error bars indicate mean ± se (n = 9). Statistical significance was determined by one-way ANOVA with multiple comparison correction by Tukey HSD. Different letters indicate significance groups (P < 0.001).

(C) Rosette size of plants as in (B); statistical significance tested as above.

(D) Localization of fluorescent TRB1 and TRB3 wild-type and mutant fusion proteins transiently produced in tobacco leaves. Bar = 10 μm.

(E) Colocalization of fluorescent TRB1, TRB3, and LHP1 fusion proteins transiently produced in tobacco leaves Bar = 10 μm.

(F) Average intensity of TRB1-GFP and LHP1-RFP across regions of interest. Error bars represent Student’s t test confidence intervals (n = 9).

(G) Average intensity of TRB3-RFP and LHP1-GFP as in (F).

(H) Telomere length analysis. DNA was prepared from Col-0, lhp1-3, trb1-1 lhp1-3, and trb3-1 lhp1-3 pools of (100 to 200) 10-d-old seedlings and 31-d-old individual plants. Material from 10-d-old ku70 and tert (G5) seedlings was included as reference.

(I) Phenotype of lhp1-3, ku70 lhp1-3, and tert lhp1-3. Bar = 1 cm.

TRB1 and TRB3 feature an N-terminal single repeat Myb domain and a C-terminal domain related to linker histones H1 and H5 (Supplemental Figure 1C). This domain structure is shared by five Arabidopsis proteins belonging to two distinct clades that formed at the origin of seed plants (Supplemental Figure 2A and Supplemental Data Set 1). The TRB clade of Arabidopsis Myb-domain proteins contains three members, which form hetero- and homomultimers (Kuchar and Fajkus, 2004). Histochemical data showed that TRB1 and TRB3 are expressed throughout the plant (Supplemental Figure 3), suggesting that they have not subfunctionalized by evolving distinct tissue-specific expression patterns. In the wild-type background, single and combined loss of TRB1 and TRB3 did not result in obvious phenotypic alterations, supporting the notion of functional redundancy between the factors (Figures 1A to 1C). By contrast, triple trb1-1 trb3-1 lhp1 mutants showed stronger lhp1 enhancement than both double mutants, indicating that TRB1 and TRB3 are only partially redundant in the sensitized lhp1 background (Figures 1A to 1C).

The trb1-1 and trb3-1 alleles changed conserved codons within the Myb domain to encode phenylalanine instead of leucine and glutamic acid instead of glycine, respectively (Supplemental Figure 2B). TRBs were previously shown to bind plant telomere repeat sequences (Schrumpfova et al., 2004; Hofr et al., 2009). Compared with the wild-type version, proteins encoded by trb1-1 fused to GFP showed a loss of fluorescence foci in the nucleoplasm (Figure 1D, green channel, two left-most panels). Based on previous colocalization studies in the same experimental system, these foci overlap with telomere repeat regions detected by DNA-fluorescence in situ hybridization (Schrumpfova et al., 2014). By contrast, an enrichment at the nucleolus was still shared between the mutant and wild-type TRB1:GFP fusions (Figure 1D, two left-most panels). Thus, localization within the nucleolus may be independent of specific DNA binding mediated by the Myb domain. TRB3-1:GFP levels were strongly reduced compared with wild-type TRB3:GFP, indicating rapid turnover of the mutated protein (Figure 1D, red channel, first and second rows). In contrast to TRB1 and TRB3, LHP1 was neither particularly enriched nor depleted at TRB1 and TRB3 foci (Figures 1E to 1G; Supplemental Figure 4). As previously reported (Libault et al., 2005), LHP1 was relatively depleted at the centers of nucleoli (Figures 1E to 1G; Supplemental Figure 4). Thus, colocalization of both TRBs with LHP1 appeared throughout the nucleoplasm, including TRB1 and TRB3 foci, while it is less probable at the nucleolus due to the relative depletion of LHP1.

Enhancement of lhp1 by Mutations in trb1 and trb3 Is Independent of Telomere Function

Successive generations of Arabidopsis plants carrying telomerase (tert) loss-of-function alleles show progressive telomere shortening, which is perceptible from the first generation on. Until generation 5, telomere shortening in tert mutants has no phenotypic consequences, but from that point on developmental aberrations accumulate, resulting in phenotypes that resemble strong PcG loss-of-function mutants (Riha et al., 2001). A T-DNA insertion allele in TRB1 resulted in a mild reduction (10 to 20%) in telomere length after five generations, but had no phenotypic effect (Schrumpfova et al., 2014).

To investigate whether enhancement of the lhp1 phenotype by trb1 and trb3 could be explained by an accelerated pace of telomere degeneration, we measured telomere length in the relevant mutants. Irrespective of their age, seedlings or mature lhp1, trb1-1 lhp1, and trb3-1 lhp1 plants of the first homozygous generation had telomeres in the wild-type length range of 2 to 5 kb. This range is different from the respectively long and short telomeres of ku70 and tert mutants of the fifth generation, but not from wild-type (Col-0) telomeres that were used as reference (Figure 1H) (Riha et al., 2001; Riha and Shippen, 2003). Thus, phenotypic enhancement in trb1-1 lhp1 and trb3-1 lhp1, which is visible from the first generation, precedes visible telomere shortening, excluding the possibility that a role of TRBs in telomere homeostasis enhances the lhp1 phenotype. In general, the PcG pathway does not seem to play a role in telomere homeostasis, since telomere length was not affected in plants with mutations in several other PRC1 and PRC2 components, although these mutants had been homozygous for several generations (Supplemental Figure 5). In addition, altered telomere length did not influence the lhp1 phenotype, since ku70 lhp1 and tert lhp1 double mutants of the first homozygous generation had a phenotype similar to lhp1 (Figure 1I).

Misexpression of Floral Meristem Identity Genes Is Enhanced in trb1-1 lhp1 Double Mutants Compared with lhp1 Single Mutants

We profiled the transcriptome of trb1-1 and lhp1 single and double mutants by RNA-seq to uncover TRB1-related functions that would explain the observed lhp1 enhancement (see Supplemental Data Set 2 for all and Supplemental Data Set 3 for an overview of differentially expressed genes). Although phenotypically indistinguishable from the wild type, the number of misregulated genes in trb1-1 was comparable to those in lhp1 single and trb1-1 lhp1 double mutants (Figure 2A). To test how misregulation correlated to PcG-mediated repression, we overlapped the gene sets with a list of H3K27me3 target genes determined by ChIP-seq analysis. PcG target genes were not overrepresented in scenarios that included the lhp1 mutant (Fisher’s test P < 2.36e-01 and P < 9.04e-02 for lhp1 and trb1 lhp1, respectively). Although this may seem unexpected, it fits well with the observation that lhp1 shows a relatively mild phenotype not in accordance with a general misregulation of PcG target genes. PcG target genes were underrepresented in the set misregulated in trb1-1 (Fisher’s test P < 9.21e-08; Figure 2B).

Figure 2.

Effect of Mutation of TRB1 on Gene Expression in Seedlings.

(A) Venn diagram of genes differentially regulated in trb1-1, lhp1-3, and trb1-1 lhp1-3 seedlings compared with wild-type Col based on RNA-seq data (biological replicates n = 3, Baggerley’s test with FDR correction on RPKM values per gene, threshold FDR < 0.05, fold change > 2). Comparison only for upregulated genes indicated in parentheses.

(B) Number of genes up- or downregulated in each genotype. Proportion of H3K27me3 target genes indicated in gray.

(C) Transcriptional clustering of misregulated genes. K-median (k = 8) clustering after median centering each gene across all samples (arbitrary units). A representative clustering result is shown.

Misregulation of 83 genes was common to both trb1-1 and lhp1 single mutants, and of these misregulated genes, 53 were also misregulated in trb1-1 lhp1 double mutants, which shared 79 genes with trb1-1 alone (Figure 2A). Although many of these commonly misregulated genes shared the direction of change observed in the wild type, the sets did not contain obvious candidate genes that would directly link to the phenotype, making it less likely that genetic enhancement was due to additive effects. Nonadditive phenotypic enhancement in the double mutant could either be attributed to genes that newly appeared in the misregulated set of trb1-1 lhp1 double mutants or to those already affected in lhp1, but increasingly so in the double mutants. Many of the 144 genes misregulated in trb1-1 lhp1, but not in lhp1, were connected to photosynthesis rather than developmental functions, which made the latter the more plausible scenario (Supplemental Data Set 3). Indeed, AG, AP3, and, to a lesser degree, SEP3 were further upregulated in trb1-1 lhp1 compared with lhp1, suggesting that they may play a role in phenotypic enhancement (Supplemental Data Set 3). We evaluated the expression of AG, AP3, SEP3, and FT directly by qRT-PCR, which corroborated that floral meristem identity genes showed increased expression in trb1-1 lhp1 compared with lhp1 (Supplemental Figure 6).

TRB1 Affects Photosynthesis-Related Genes Alone and Developmental Regulatory Genes Together with LHP1

Since the quality of the gene set misregulated in trb1-1 alone seemed distinct from the one likely linked to phenotypic enhancement of lhp1, we performed further transcriptional pattern and Gene Ontology (GO) term enrichment analysis to uncover functional connections among groups of misregulated genes (Figure 2C; Supplemental Data Sets 3 and 4). Preliminary analysis showed that median normalization and the number of k = 8 lead to robust k-clustering results. Three of the clusters (1, 6, and 8) were predominantly affected by the lhp1 mutation with cluster 1 showing anticorrelation to clusters 6 and 8. Misexpression in cluster 6 was most obviously enhanced in the trb1-1 lhp1 double mutant background. Two clusters (2 and 5) were predominantly affected in the trb1-1 background, showing misregulation in the opposite direction. In clusters 3, 4, and 7, expression in the wild type was most distinct from all mutant backgrounds, indicating that gene expression was affected by trb1-1 and lhp1 in similar direction.

The clusters strongly diverged in their average expression levels as estimated by average read count (Figure 3A). PcG target genes were overrepresented in clusters with low and virtually absent in clusters with high read count (Figure 3B; Supplemental Data Set 3). In conclusion, genes affected by mutation of TRB1 belonged to two distinct groups. Highly expressed genes were affected by trb1-1 as a single mutation (clusters 2 and 5), while increased upregulation of a lowly expressed subset containing many PcG target genes was observed in combination with lhp1 (cluster 6).

Figure 3.

Analysis of Gene Clusters Affected in the trb1 Mutant Background.

(A) Box plot showing median normalized log-transformed expression across all genes per cluster (cl). Thick horizontal lines represent the median, boxes represent the 25th to 75th percentiles, the whiskers represent the 5th and 95th percentiles, and dots indicate outliers.

(B) Proportion of H3K27me3 target genes per cluster. Genome average of H3K27me3 is indicated by dashed line. Significant deviation from genome average was tested by Fisher’s exact (*P < 0.05, **P < 0.01, and ***P < 0.005).

(C) Gene clusters were further clustered according to the number of shared GO terms (dendrogram). Curved lines indicate positive (+) and negative (−) correlations between expression of the clusters. The most significant shared GO term is indicated below the tree. The pictogram indicates statistical significance of enriched GO term (yellow, lower; orange, intermediate, red, higher significance). An enriched site II sequence motif found in cluster 5 is indicated below.

We used hierarchical clustering to further analyze whether the k groups shared enriched GO terms (Figure 3C; Supplemental Data Set). Functional connections were most significant for clusters 2, 5, and 7 with shared GO term enrichment for chloroplast and photosynthesis functions. In addition, the anticorrelated clusters 1 and 3 overlapped in pathways resulting in the formation of organic and inorganic nitrogenous compounds, which include tetrapyrrols. Clusters 6 and 8, which showed lhp1-dependent upregulation, were more loosely connected by higher level terms, such as response to hormone stimulus and inorganic substance. Only cluster 8 was strongly enriched for GO terms belonging to flower development, although some of the key genes attributed to the lhp1 phenotype such as AG and AP3 were found in cluster 6.

Site II TCP Binding Motif Is Related to Reduced Expression in the trb1-1 Mutant

We searched for enriched sequence motifs within each k-cluster using 500 nucleotide proximal promoter regions. Long A-tracts were enriched in all clusters but cluster 5 was also enriched for a putative cis-element (AGGCAAA), previously described as site II TCP binding motif (Tremousaygue et al., 2003; Welchen and Gonzalez, 2005). In combination with a minimal promoter, the site II motif had been shown to be necessary and sufficient to drive expression in rapidly cycling cells (Tremousaygue et al., 2003). Interestingly, the site II motif is often associated with teloboxes in promoters of genes encoding ribosomal proteins and components of the translational apparatus. Teloboxes (AAACCCTA) are related to telomeric repeats (CCCTAAA × n; n = 2-1000+), which were previously shown to bind TRB1 in vitro (Schrumpfova et al., 2004; Hofr et al., 2009). The presence of teloboxes enhances the effect conferred by site II motifs but is not sufficient to drive expression from a minimal promoter on its own (Tremousaygue et al., 2003; Gaspin et al., 2010). Cluster 5 was indeed enriched for ribosomal genes, corroborating the previous link to site II motifs. However, teloboxes, although present, were not overrepresented in the corresponding promoter set, which could be due to the high incidence of teloboxes among all Arabidopsis promoters used as a background in the enrichment analysis (Tremousaygue et al., 2003; Gaspin et al., 2010). Cluster 5 contained the most highly expressed gene set (Figure 3A), which was expressed at lower levels in the trb1-1 and trb1-1 lhp1 background. Thus, within cluster 5, TRB1 binding to teloboxes would fit the previously described model of enhanced upregulation of site II containing promoters by the presence of TRB1 bound to teloboxes.

TRB1 Binds to Thousands of Sites

While the enrichment of site II motifs at genes repressed in the trb1-1 backgrounds suggested that misregulation in cluster 5 was a direct effect, it did not explain enhanced upregulation of PcG target genes in cluster 6. We therefore performed ChIP-seq experiments in Arabidopsis seedlings to identify direct target genes of TRB1.

ChIP-seq libraries were prepared in two biological replicates using chromatin prepared from CaMV Pro35S:TRB1:GFP lines in the trb1-1 background (from now TRB1:GFP). The lines were generated by segregating the lhp1 mutation away from a complemented TRB1:GFP lhp1 line (Supplemental Figure 1A). TRB1:GFP significantly enriched 7825 genomic sites over a background prepared from wild-type chromatin and precipitated with the same antibodies (SICER pipeline, false discovery rate [FDR] < 0.0001; Supplemental Figure 7, Supplemental Data Set 5, and https://gbrowse.mpipz.mpg.de/cgi-bin/gbrowse/arabidopsis10_turck_public/). The overlap of target regions between biological replicates was between 73 and 80% and a majority were around 250 bp long, indicating single or closely clustered binding events (Supplemental Figure 7B).

Sequences under the TRB1 peaks were scored for enrichment and for their probability of being located at the center of enriched regions. The telobox was among the most significantly enriched motifs (MEME, P < 2.84.6e-61; Figure 4A; Supplemental Data Set 6 and Supplemental Figure 8). The analysis also revealed a previously undescribed related CRACCTA motif, now named celobox (DREME, P < 5.8e-96), that was even more strongly enriched at peak centers (Figure 4A). In fact, by far the most significant enrichment was detected for a shorter telobox-related motif (RMCCTA) that is included in both telo- and celoboxes (DREME, P < 1.3e-273; Figure 4A). In addition, the consensus sequence RGCCCW, which comprises the site II motif was significantly enriched, although positioned slightly off-center (Figure 4A).

Figure 4.

Analysis of TRB1 Target Sites in Seedlings.

(A) cis-elements enriched in TRB1 target regions. Regions enriched in TRB1:GFP compared with Col were interrogated for centrally enriched cis-elements by MEME-ChIP. Consensus sequences with corresponding E-values (bottom) and probabilities of occurrence around the fragment’s center (top).

(B) Proportion of TRB1 targets among the genes encoding ribosomal functions. 1, Ribosomal protein; 2, tRNA; 3, intron-containing tRNA; 4, snoRNA.

(C) Proportion of H3K27me3 target (blue) and non-target (red) genes for TRB1 positive (upper pie) and negative (lower pie) genes.

(D) and (E) Metagene analysis of TRB1 target genes.

(D) Background-corrected TRB1:GFP ChIP-seq reads for all genes (green), all TRB1 target genes (brown), and H3K27me3 negative (purple) or positive (pink) TRB1 target genes.

(E) Frequency of RMCCTAR consensus for gene categories as in (D). Genes are represented in relative length from TSS to transcriptional exit site (TES). Sequences 5′ and 3′ of TSS and transcriptional exit site, respectively, are scaled in base pairs.

To evaluate the chromatin context in which TRB1 binding sites were located, we used data from a study that classified the genome of the Arabidopsis seedling into nine distinct chromatin types based on a comprehensive analysis of genome-wide chromatin modification data (Sequeira-Mendes et al., 2014). We compared chromatin states at TRB1 binding sites to 100 permutated data sets with randomly reshuffled binding coordinates to test for significant over or underrepresentation. Chromatin states 1 and 2 are enriched for proximal promoter signatures such as H3K4me2, H3K4me3, H3 acetylation, H3K36me3, H2Bub, and H2A.Z but differ by the respective absence and presence of H3K27me3. Both states were significantly overrepresented in TRB1 peaks, while states 7, 8, and 9, corresponding to intergenic regions and heterochromatin signatures, are clearly underrepresented. Likewise, state 3 enriched in marks for elongating transcription is underrepresented (P < 0.01). State 4 is similar to state 2, but without the marks corresponding to transcription start sites (TSSs). H3K27me3 levels are high in state 4, but distal promoter regions rather than gene bodies are represented. State 4 is overrepresented among the TRB1 target regions (P < 0.01). States 5 and 6, corresponding to intergenic regions with or without H3K27me3, respectively, are neither over- nor underrepresented (Supplemental Figure 9).

We scored 6782 genes as TRB1 target genes by overlapping enriched regions with either the promoter (up to 3 kb from the ATG) or the gene body (including 250-bp downstream regions). Consistent with previous reports on the coincidence of site II motifs and teloboxes (Tremousaygue et al., 2003; Gaspin et al., 2010), we found ribosomal protein genes overrepresented among direct TRB1 target genes (Figure 4B). In addition, while tRNA and snoRNA genes were marginally enriched among the targets, a subset of intron-containing tRNA genes was strongly overrepresented (Figure 4B).

Only k-cluster 1 and, as expected, cluster 5 showed an overrepresentation of direct TRB1 target genes (Fisher’s exact test, P < 3.30e-02 and P < 6.83e-03, respectively). By contrast, cluster 6, which is H3K27me3 enriched and most affected with regard to expression in trb1 lhp1 double mutants, was depleted in target genes, although the effect was not statistically significant. However, only 14% of TRB1 target genes were also PcG targets, which is significantly less than the expected genome-wide 20% (Fisher’s exact test, P < 2.00e-08; Figure 4C).

Strikingly, the TRB1 binding pattern across target loci differed for PcG target and non-target genes. While TRB1 almost exclusively bound at the TSS of genes that were not PcG targets, it associated more evenly across gene bodies at PcG target genes (Figure 4D). Occurrence patterns of telobox-related motifs recapitulated this differential distribution by being particularly enriched across the gene bodies of H3K27me3-positive TRB1 target genes, while most strongly enriched at the promoters of the other genes (Figure 4E).

In conclusion, TRB1 binds preferentially to telobox-related motifs located at the TSS for most of its targets, while at PcG target genes, TRB1 binding spreads across gene bodies and, based on its relative enrichment in chromatin state 4, possibly also distal promoter regions. This pattern is mirrored by an increased occurrence of telobox-related motifs.

LHP1 Competes with TRB1 at PcG Target Genes

It seemed likely that LHP1 prevents TRB1 from binding to target sites located at PcG target genes, which would be congruent with the observation that effects on expression were only observed in the trb1-1 lhp1 but not the trb1-1 background. We tested this by performing ChIP-seq experiments with chromatin prepared from TRB1:GFP lhp1 seedlings. The levels of TRB1:GFP as well as its accumulation in the nucleus and nucleolus were unaffected by the lhp1 mutation (Supplemental Figure 10). Nevertheless, the number of target regions passing the enrichment level was significantly increased in the lhp1 background, although H3K27me3 target genes were still underrepresented (Fisher’s exact test P < 2.20e-16; Figure 5A; Supplemental Data Set 7). Increased binding of TRB1 in lhp1 was quantitative more than qualitative, since average signals at target sites were slightly higher for regions already enriched in the wild type, while lhp1 unique regions often failed to pass the stringent significance threshold for enrichment in the wild type (Figure 5B). Increased TRB1 binding was detected at both H3K27me3-positive and -negative binding sites (Supplemental Figure 11). AG, AP3 (cluster 6), and SEP3 (cluster 8) were among the TRB1:GFP lhp1 target genes, but only SEP3 was already scored as positive in TRB1:GFP (Figures 5C and 5D; https://gbrowse.mpipz.mpg.de/cgi-bin/gbrowse/arabidopsis10_turck_public/). Likewise, TRB1 bound the FT (cluster 8) promoter regions in both backgrounds (Supplemental Figure 12). We confirmed these results by ChIP-qPCR for AG, AP3, SEP3, and FT (Supplemental Figure 12).

Figure 5.

Comparison of TRB1 Target Sites in the Wild Type and lhp1.

(A) Venn diagram showing the number of regions associated with TRB1:GFP (gray) and TRB1:GFP lhp1 (red).

(B) Read depth (corrected for read depth at control precipitation) across all fragments enriched only in TRB1:GFP lhp1 (left panel) or in both genetic backgrounds (right panel). ChIP-seq data are based on two biological replicates of TRB1:GFP (black lines) and TRB1:GFP lhp1 (red lines). Enriched fragments are displayed between gray lines on a fraction of length scale and flanking regions on a base pair scale.

(C) Overview of the SEP3 locus. Top panels show gene models with exons and introns illustrated by boxes and lines, respectively. Untranslated regions are depicted by lighter blue fill color; direction of the coding strand is indicated by the arrow. Location of telobox-related, telobox, and celobox motifs are indicated by blue, red, and green boxes, respectively. Middle panels show coverage of TRB1:GFP and TRB1:GFP lhp1 corrected by coverage from Col control precipitation. Values more than 50 reads over background are indicated in black for TRB1:GFP and red for TRB1:GFP lhp1. Black and red boxes indicate location of fragments indicates significantly enriched by SICER (FDR < 0.0001). Two bottom panels show ChIP-chip enrichment of LHP1:HA and H3K27me3 from our previously published data (Dong et al., 2012; Engelhorn et al., 2012).

(D) Overview of AG locus; colors and symbols same as for (C).

(E) Overview of the AP3 locus; colors and symbols same as for (C).

In general, the increase in the number of target genes was proportional to the genome average for all transcriptional clusters, with the notable exception of cluster 6 that went from underrepresented in TRB1 targets in the wild type to enriched in lhp1 compared with the genome-wide distribution, although neither value was statistically significant (Figure 6A). However, this relative enrichment became highly relevant if targeting by both H3K27me3 and TRB1 was considered (Figure 6B).

Figure 6.

Comparison of TRB1 Target Genes within Transcriptional Cluster Groups between the Wild Type and lhp1.

(A) Scatterplot of proportion of TRB target genes per transcriptional cluster for the wild type and lhp1 among genes misregulated in trb1-1, lhp1, and trb1-1 lhp1. Genome-wide proportion for the wild type and lhp1 is indicated by a black and a red line, respectively. Statistical difference from expected indicated by asterisks (Fisher’s exact test, *P < 0.05, **P < 0.01, and ***P < 0.005).

(B) Proportion of genes that are both TRB1 and H3K27me3 positive in the wild type and lhp1. Genome-wide proportion indicated by a black and a red line for the wild type and lhp1, respectively. Significant deviation from genome average was tested by Fisher’s exact (*P < 0.05, **P < 0.01, and ***P < 0.005).

In conclusion, LHP1 prevents binding of TRB1 at many target sites, which are not all scored as PcG-positive genes. Increased binding of TRB1 to PcG target genes occurs in particular at transcriptional cluster 6, which is the most upregulated in the trb1-1 lhp1 double mutant.

DISCUSSION

A Model for the Role of TRB1 in Transcriptional Regulation

Here, we report that TRB1 can affect the expression of direct target genes by binding to telobox-related elements. The direction in which expression is altered in comparison to relevant controls cannot be predicted, although the following tendencies were detected. First, a functionally correlated set of highly expressed TRB1 target genes is expressed at lower levels in the absence of TRB1 in both the wild type and lhp1 mutant background. A second group, corresponding predominantly to PcG target genes showing TRB1 binding only in the lhp1 mutant background, shows strongly enhanced induction in trb1-1 lhp1 compared with lhp1. At these loci, TRB1 seems to implement a second layer of repression.

Several models can explain our findings (Figure 7). First, it seems likely, based on previous data showing the synergistic role of teloboxes and site II motifs (Tremousaygue et al., 2003), that TRB1 binding to teloboxes per se is neutral to transcription but assists in the activation function realized by TCP factors binding to site II. It has been reported that transcription factor binding sites need to cluster, as binding of a single one would not be sufficient to displace nucleosomes from cognate binding sites (Mirny, 2010; Moyle-Heyrman et al., 2011). In such a scenario, the presence of TRB1 would assist in liberating the binding site for TCPs or other transcription factors that could either be transcriptional activators or repressors. In a more elaborate model, TRB1 binding would not be quite neutral but lead to the recruitment of chromatin remodeling factors that promote binding of other factors. The role of TRB1 at PcG target genes could be explained by the participation of chromatin remodelers because these can participate in maintaining a less permissive chromatin structure. Such ambivalent roles as activators and repressors have been reported for the complex around the CHD4-type ATPase PICKLE (PKL), which reportedly antagonizes PcG-mediated repression (Aichinger et al., 2011), but at the same time helps to maintain H3K27me3 at several direct target genes (Zhang et al., 2012). Likewise, the SWI/SNF2-type ATPase BRAHMA is required to repress the PcG target gene FLC (Farrona et al., 2011a) while antagonizing H3K27me3 deposition at many of its direct targets (Li et al., 2015). As for the last example, the SWR1 complex that exchanges canonical histone H2A with H2A.Z is thought to facilitate both transcriptional activation and repression by creating boundaries for transcription factor binding sites at TSS (Kumar and Wigge, 2010; Farrona et al., 2011a).

Figure 7.

Working Model for Transcriptional Regulation by TRB1.

(A) and (B) Model for highly expressed genes in cluster 5.

(A) TRB1 binds to telobox-like motifs, thereby facilitating the binding of TCP factors to site II motifs. Genes are highly expressed in the absence of TRB (ON), while the presence of both factors helps to express genes at very high levels (ON+), but the presence of TCP is a dominant requirement for expression (OFF). In principle, the mechanisms could also apply to coregulation by other cis-elements and cofactors, which could also be repressors.

(B) As in (A) but TRB1 recruits chromatin remodelers to assist binding of TCP factors or support their downstream effect on transcription activation.

(C) and (D) TRB1 action at PcG target genes, showing enhanced upregulation in the double mutant.

(C) TRB1 basically functions as in (A) but facilitates the action of a repressor that participates in downregulation but has no dominant effect. In the wild type, there is an equilibrium between LHP1-dependent and TRB1-dependent repression and targets are always repressed (OFF). In lhp1 plants, a repressor can bind with the help of TRB1 and attenuate upregulation (ON). In trb1 lhp1 double mutants, attenuation is lost, leading to enhanced expression (ON+).

(D) TRB1 acts similar as in (B) by recruiting a chromatin remodeler. The remodeler maintains a more closed chromatin conformation but can only partially compensate for the lack of LHP1. Genes are not expressed in the wild type and trb1 single mutants (OFF), are induced in lhp1 mutants (ON), and are hyperinduced in trb1 lhp1 double mutants (ON+).

Telobox-Related Elements and Transcription in an Evolutionary Context

The presence of cis-elements related to telomere repeats is conserved in plants, yeast, and human (Ruiz-Herrera et al., 2008; Vaquero-Sedas and Vega-Palas, 2011). It has been speculated that these evolve rapidly from interstitial telomere repeats, which form by the invasion of chromosomal ends into extratelomeric regions (Gaspin et al., 2010). In any case, it seems that the connection of telomere-derived cis-elements and cognate transcription factors with genes encoding proteins involved in primary cellular functions is an ancient one, although the outcome on gene regulation can be quite different from the one reported in plants (Ye et al., 2014). For example, the yeast Myb-related factor Rap1 binds at telomeric and extratelomeric sites, where it can contribute to either up- or downregulation of target genes. During senescence, the shortening of telomeres releases more Rap1 protein to bind at extratelomeric sites, leading to downregulation of genes encoding core histones and the translational apparatus, while supporting upregulation of other target genes responsive to senescence (Platt et al., 2013). In mammals, the TELOMERE REPEAT FACTOR2, a homolog of plant TRBs, binds to extratelomeric sites, which can contribute to either up- or downregulation of target genes (Yang et al., 2011).

cis-Elements at PcG Target Genes

At PcG target genes, TSS sites were present at a normal proportion but were not as likely bound by TRB1 as were other target genes. By contrast, binding across the gene body was increased, as was the number of telobox-related motifs. It is possible that the increased number of motifs is a means to compensate for the more restricted access due to chromatin compaction at H3K27me3-positive genes. It is possible that these sites become accessible during transcription, which may occur at a low rate due to spontaneous chromatin state switching (Angel et al., 2011; Satake and Iwasa, 2012). In this case, the presence of TRB1 as a second layer of repression is important to avoid leaky expression of target genes.

Our study is not the first that links the PcG pathway to telobox motifs. A genome-wide survey of binding sites for the PRC2 component FERTILIZATION INDEPENDENT ENDOSPERM (FIE) reported an enrichment of teloboxes as well as GAGA and GAAGAA repeat motifs at FIE-enriched peaks (Deng et al., 2013). Furthermore, a recent study reports an overrepresentation of teloboxes particularly at those PcG target genes that show strongly reduced H3K27me3 levels in the clf mutant background (Wang et al., 2016). A link between the GAGA-motif and PcG recruitment was suggested by showing that GAGA binding proteins, such as BPC6, could bind to both GAGA-motifs and LHP1 in vitro (Hecker et al., 2015). In the case of TRBs, we think it unlikely that they act as recruiters of PcG components due to their extensive binding at PcG non-target genes.

Conclusion

We find that TRBs are novel transcriptional coregulators that potentially impact thousands of genes, a number of which are more commonly associated with chromatin regulatory complexes than sequence-specific transcription factors, which show more restricted binding. TRBs seem to assist rather than define target gene regulation. In consequence, their presence does not predict the direction of transcriptional regulation, which is defined by the presence of other cis-elements or chromatin components. At PcG target genes, TRB1 binding is generally reduced, while the number of cognate binding sites increases. It seems that PcG complexes rely on TRBs as second-layer repressive backup.

METHODS

Mutagenesis, Mutant Screening, and Cloning

Mutagenesis, screening for enhanced lhp1 genotypes, and cloning of causal mutations was as previously described (Hartwig et al., 2012). Briefly, 200 mg of lhp1-3 (Larsson et al., 1998) mutant seeds was incubated in 100 mL of 30 mM EMS for 12 h after pretreatment in 0.1% KCl solution at 4°C for 14 h. Mutagenized seeds were washed with distilled water and incubated in 100 mL sodium thiosulfate (100 mM, 15 min) followed by three washing steps in 500 mL of deionized water 30 min prior to transfer to soil.

For mutant screening, M2 mutant families were grown in soil in a greenhouse at 22°C in short-day conditions (8 h light/16 h dark) and scored visually for an enhancement of the early flowering and reduced size lhp1 mutant phenotype. M3 seeds of mutants were grown in growth chambers for confirmation of the phenotype (60% humidity, 12 h; 16°C light/12 h dark; 14°C cycles).

For cloning of the causative gene, phenotypically confirmed M3 mutants were backcrossed twice to lhp1-3 to generate a BC2F2 population of 1000 individuals. Leaf material from plants with a mutant phenotype (trb1 lhp1, n = 240; trb3 lhp1, n = 295) was pooled for DNA preparation (DNeasy Plant Maxi Kit; Qiagen; according to the manufacturer’s instructions) and NGS library preparation (Illumina True-seq). On the IlluminaHiSeq platforms 60E6 and 80E6, 50-bp reads were generated for trb1 lhp1 and trb3 lhp1, respectively, resulting in genome coverage of 37- and 50-fold. Using the fast isogenic mapping pipeline (Hartwig et al., 2012), four linked possibly homozygous candidate single nucleotide polymorphisms resulting in nonsynonymous codons were identified per genotype. Of these, only trb1-1 and trb3-1 alleles represented homologs.

Plant Materials

The EMS-induced lhp1-3 allele in the Col background has been described previously as terminal flower 2-1 (tfl2-1) but has also been referred to as tfl2-2 (Kotake et al., 2003). The trb1-2 and trb3-2 alleles were obtained from the SALK T-DNA insertion line collection (SALK_001540 and SALK_134641, respectively). The ku70, tert (−/+), and tert (G5) mutants were provided by Karel Riha at the Gregor Mendel Institute of Molecular Plant Biology, Austria.

Cultivation Conditions

For qRT-PCR/RNA-seq and ChIP-PCR/ChIP-seq, seeds of corresponding Col-0 and mutants were sterilized in 75% ethanol and sown on GM medium. Materials from 10-d-old seedlings grown in Percival growth chambers at 22°C in long days (16 h light/8 h dark) were collected. For phenotypic analysis, seeds were sown on soil and transferred to long days after stratification (4°C, 3 d). Flowering time was determined by counting the number of rosette and cauline leaves of the main shoot. Plant size was measured as longest diameter at bolting. Test for statistical significance was performed by one-way ANOVA followed by a Tukey HSD correction for multiple comparisons and by a Student’s t test for comparisons of two groups.

Plasmid Construction and Generation of Transgenic Plants

Full-length LHP1, TRB1, and TRB3 cDNAs without stop codons were amplified from Col-0 cDNA. Genomic sequence (2.0 kb upstream of ATG, full gene body, and 1.5 kb downstream of stop codon) and promoter sequence (2.0 kb upstream of ATG) of TRB1 and TRB3 were amplified from genomic DNA of Col-0. Oligonucleotide primers were Gateway (GW) compatible and are indicated in Supplemental Table 2. Fragments were introduced into the pDONR207 vector (Invitrogen) and then to Agrobacterium tumefaciens binary vectors by GW recombination reactions according to the manufacturer’s instructions. Coding sequences of LHP1, TRB1, and TRB3 were recombined into Pro35S:GW:GFP-pAM or Pro35S:GW:tagRFP-pCZN654 (a gift from Richard Immink) to create C-terminal fusions to GFP and RFP under the control of the CaMV 35S promoter; genomic TRB1 and TRB3 sequences were introduced into pGD2B (a gift from Hailong An) to allow expression under the control of their native regulator regions. TRB1 and TRB3 promoter sequence were introduced into GW:GUS-pGREEN (Adrian et al., 2010) to drive expression of the GUS reporter gene. Vector backbones are provided in GenBank format (Supplemental Data Set 8). Transgenic plants were generated by Agrobacterium-mediated transfer using the floral dip method (Clough and Bent, 1998). TRB1:GFP expression constructs were generated in the trb1-1 lhp1 mutant background from which the lhp1-3 allele was subsequently removed by crossing to Col-0 wild type.

Fluorescence Microscopy

To determine the cellular localization of LHP1, TRB1, and TRB3, fusions to GFP and tagRFP were transiently expressed in 3-week-old Nicotiana benthamiana (tobacco) plants by Agrobacterium infiltration. Briefly, Agrobacterium strains carrying plasmids encoding fusion proteins and the p19 silencing suppressor were grown overnight at 28°C in 10 mL YEP medium plus selective antibiotics and then collected and resuspended in infiltration buffer (10 mM MgCl₂, 10 mM MES, pH 5.6, and 150 μg/mL acetosyringone). Resupended bacteria were incubated at 28°C for 3 h in the dark and infiltrated to the lower surface of tobacco leaves with a needle-free syringe. Cellular localization of fusion proteins was examined under a LSM 700 confocal laser scanning microscopy (Carl Zeiss). Comparative fluorescence intensity scanning was performed by pixel density analysis using ImageJ software.

Terminal Restriction Fragment Analysis

DNA was extracted from pooled (10 to 20) 10-d-old seedlings and 31-d-old single plants with the DNeasy Plant Mini Kit (Qiagen). DNA (3 µg) was digested with restriction endonuclease Tru1I (Fermentas) at 42°C overnight. Digested DNA was electrophoresed on an agarose gel and blotted to a polyvinylidene fluoride membrane. (T₃AG₃)₄ oligonucleotide was end-labeled by ³²P and was used as hybridization probe for DNA gel blotting.

GUS Staining

For GUS staining, 10-d-old seedlings were incubated for 30 min in 90% (v/v) acetone on ice, rinsed with 50 mM sodium phosphate buffer, pH 7.0, and incubated overnight at 37°C in staining solution (0.5 mg/mL X-Gluc, 50 mM sodium phosphate buffer, pH 7.0, 0.5 mM potassium ferrocyanide, 0.5 mM potassium ferricyanide, and 0.1% [v/v] Triton X-100). After staining, samples were washed with 50 mM sodium phosphate buffer, pH 7.0, and cleared in 70% (v/v) ethanol. The GUS staining results were visualized under a light stereomicroscope (MZ 16 FA; Leica).

RNA Isolation, Quantitative RT-PCR, and RNA-Seq

Total RNA was extracted using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. Total RNA (5 µg) was treated with DNaseI (DNA-free kit; Ambion). For RT-PCR, cDNA was generated at 42°C for 2 h using Superscript II reverse transcriptase and T18 oligonucleotide for priming (Life Technologies). Expression of TRB1 and TRB3 of T-DNA lines was measured by PCR using PP2A as control. For quantitative RT-PCR, experiments were performed in a Bio-Rad iQ5 apparatus using a homemade Eva-GREEN amplification cocktail (80 mM KCl, 20 mM Tris-HCl; pH 8.0, 5 mM MgCl₂, 0.4 mM deoxynucleotide triphosphate, 400 nM forward and reverse oligonucleotide primer, 1× EVAGreeen dye [Biotium], and 0.1 units/μL Taq polymerase) for detection. Quantification was performed using the relative -ΔΔCT method using PP2A as reference. Oligonucleotide primers are indicated in Supplemental Table 2. For RNA-seq, material was collected from three independent biological replicates, and DNA-free total RNA was generated as described above. Illumina True-seq library preparation was performed from DNA-free total RNA (3 µg) by the Max Planck Genome Centre in Cologne. Quality trimmed single-end RNA-seq reads were mapped to the Arabidopsis TAIR10 annotation using the CLC genomics workbench (parameters for quality trimming: removal of reads with ambiguous nucleotides >2, quality score <0.05, length <80 nucleotides; parameters for mapping: mismatches <3). Read per kilobase of exon models (RPKM) were calculated from 8.3e6 – 11.4e6 mapped reads per library. RPKM values were scaled to fit the medians of each library and differential expression determined using Baggerley’s weighted Z-test with FDR correction as implemented by the CLC genomics workbench (threshold FDR < 0.05, fold change > 2). Transcriptional cluster analysis was performed using the Cluster3.0 software package after centering average RPKM values per gene to the median value across all samples (de Hoon et al., 2004). After empirical evaluation, k = 8 was selected for k-median clustering using Eucledian distance. Cluster results were visualized using Java TreeView (Saldanha, 2004). GO analysis was performed using the AgriGO Web tool (Du et al., 2010). Enriched GO terms were compared between k-clusters using AgriGO compare.

ChIP and ChIP-Seq

ChIP experiments were performed as previously described (Reimer and Turck, 2010). Chromatin was extracted from 10-d-old whole seedlings (1 to 3 g). GFP (Abcam; Ab290) and H3K27me3 (Millipore; 07-449) antibodies were used for chromatin immunoprecipitation. For ChIP-PCR, amplification was performed in a Bio-Rad iQ5 apparatus using a home-made Eva-GREEN amplification cocktail (80 mM KCl, 20 mM Tris-HCl, pH 8.0, 5 mM MgCl₂, 0.4 mM deoxynucleotide triphosphate, 400 nM forward and reverse oligonucleotide primer, 1× EVAGreeen dye [Biotium], and 0.1 units/μL Taq polymerase) for detection or Bio-Rad SYBR Green master mix. TRB1:GFP binding and H3K27me3 enrichment were normalized to input DNA prepared from a reverse cross-linked aliquot of each chromatin preparation. Quantitative PCR data are shown as the means of three technical replicates from a representative experiment from at least two biological replicates. Primers used for ChIP-PCR are shown in Supplemental Table 2. For ChIP-seq, two immunoprecipitations from independent biological replicates were processed for NGS library preparation. All libraries were prepared by the Ovation Ultralow Library Systems (NuGEN) following the manufacturer’s instructions using 80% of a typical ChIP as starting material. After amplification for 16 PCR cycles, DNA of a size range of between 200 and 300 bp was purified from an agarose gel. Amplification was confirmed by testing an aliquot of the library before and after amplification by quantitative PCR. Sequencing was performed as single-end 100-nucleotide reads on an Illumina HiSeq by the Max Planck Genome Centre in Cologne. ChIP-seq reads were mapped to the TAIR10 assembly of Arabidopsis thaliana using the CLC genomics workbench. To identify TRB1 target regions, reads with a number of mismatches of >2 and more than one mapping position were discarded. Only one of two or more identical reads was kept for further analysis, resulting in between 12.8e6 and 28.0e6 mapped reads per sample. Enriched regions were determined from each sample by SICER (Zang et al., 2009) using libraries prepared from Col anti-GFP mock ChIP as background, windows of 80 bp, and a FDR < 0.0001 as threshold. Overlapping and directly adjacent enriched windows were merged to enriched regions. For annotation to target genes, the annotatePeaks.pl function of the Homer suite was used (Heinz et al., 2010). A custom annotation file was prepared based on the Arabidopsis TAIR10 annotation that allowed assigning enriched regions first to TSS and transcriptional exit sites ±250 bp, then to gene body regions, then to 1-kb promoter regions, and last to 3-kb promoter regions. Each fragment was assigned to only one gene. For detection of H3K27me3-enriched target regions, mapping was performed as above except that reads with more than one mapping position were randomly distributed among mapping sites. Enriched regions were detected by SICER as described above except that window size = 300 bp and FDR < 0.01 were used and the background was determined based on pooled input reads from four independent libraries prepared from reverse cross-linked chromatin. Enriched windows were merged to enriched regions if their distance was below 600 bp. Finally, enriched regions were intersected between replicates using bedtools (Quinlan and Hall, 2010) and the intersections used for further analysis. Genes were annotated as H3K27me3 positive if at least 80% of their gene body overlapped with a H3K27me3-positive region. Read traces for gbrowse were produced by randomly selecting 11e6 and 12e6 mapped reads from each TRB1 and H2K27me3 sample, respectively, extending the reads to 500 bp and calculating read depth per genomic position using bedtools (Quinlan and Hall, 2010). Coverage depth from two biological replicates was summarized to visualize averages using bedtools.

Motif Enrichment and Metagene Analysis

Enriched motifs in TRB1 target regions were identified using the MEME-ChIP Web tool (Machanick and Bailey, 2011). For region intersection and sequence extraction, the bedtools suite (Quinlan and Hall, 2010) was used. For metagene analysis, the TAIR10 assembly was annotated for enriched motifs using EMBOSS function fuzznuc (Rice et al., 2000). Metagene analysis was performed using ngs.plot.r (Shen et al., 2014). For this purpose, annotation data were converted to the binary.bam file format using custom scripts and bedtools.

Accession Numbers

Read data for RNA-seq and ChIP-seq experiments are accessible at EBI under accession code ERA422470. Processed ChIP-seq data can be visualized at the following public gbrowse link: https://gbrowse.mpipz.mpg.de/cgi-bin/gbrowse/arabidopsis10_turck_public/. Gene annotation data from this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: TRB1 (AT1G49950), TRB3 (AT3G49850), LHP1 (AT5G17690), AG (AT4G18960), SEP3 (AT1G24260), AP3 (AT3G54340), FT (At1g65480), and PP2A (AT1G13320).

Supplemental Data

• Supplemental Figure 1. Confirmation of mutant mapping results.

• Supplemental Figure 2. Phylogenetic analysis of single-myb-Histone1/5 domain proteins in plants.

• Supplemental Figure 3. Histochemical detection of GUS activity for TRB1 or TRB3.

• Supplemental Figure 4. Representative comparative fluorescence profile of coexpressed TRB1 and TRB3 with LHP1 in tobacco cells.

• Supplemental Figure 5. Telomere length in PcG pathway mutants.

• Supplemental Figure 6. Quantitative RT-PCR analysis to confirm expression levels of TRB1 target genes in lhp1-3 and trb1-1 lhp1-3 background.

• Supplemental Figure 7. Replicate ChIP-seq analysis of TRB1 target regions.

• Supplemental Figure 8. Full analysis of cis-elements enriched under TRB1 ChIP-seq enriched regions in wild-type and lhp1 backgrounds.

• Supplemental Figure 9. Overlap of TRB1 binding with chromatin topologies.

• Supplemental Figure 10. TRB1-GFP localization and level in wild-type and lhp1 backgrounds.

• Supplemental Figure 11. Differences in TRB1 binding strength at binding sites in wild-type and lhp1 backgrounds.

• Supplemental Figure 12. TRB1 binding at FT and ChIP-PCR confirmation of results for AP3, SEP3, AG, and FT.

• Supplemental Table 1. Table of SNPs identified by isogenic mapping-by-sequencing.

• Supplemental Table 2. List of oligonucleotides used in this study.

• Supplemental Data Set 1. Sequence alignment in fasta format.

• Supplemental Data Set 2. Comprehensive table of RNA-seq results.

• Supplemental Data Set 3. Summary table of RNA-seq and ChIP-seq results for differentially expressed genes.

• Supplemental Data Set 4. Summary of GO term enrichment analysis, ready for visualization with AgriGO custom compare tool.

• Supplemental Data Set 5. Table of genomic fragments bound by TRB1:GFP in both replicates prepared from the trb1-1 background and annotation of target genes.

• Supplemental Data Set 6. Position weight matrices for cis-elements enriched under TRB1 and TRB1 lhp1 ChIP-seq peaks.

• Supplemental Data Set 7. Table of genomic fragments bound by TRB1:GFP in both replicates prepared from the trb1-1 lhp1 background and annotation of target genes.

• Supplemental Data Set 8. Plasmid backbones in GenBank format.

Glossary

ChIP-seq

chromatin immunoprecipitation sequencing

EMS

ethyl methanesulfonate

GO

Gene Ontology

FDR

false discovery rate

TSS

transcription start site

RPKM

read per kilobase of exon models