The profile of repeat-associated histone lysine methylation states in the mouse epigenome Joost HA Martens1, Roderick J O���Sullivan1, Ulrich Braunschweig, Susanne Opravil, Martin Radolf, Peter Steinlein and Thomas Jenuwein* Research Institute of Molecular Pathology (IMP), The Vienna Biocenter, Vienna, Austria Histone lysine methylation has been shown to index silenced chromatin regions at, for example, pericentric heterochromatin or of the inactive X chromosome. Here, we examined the distribution of repressive histone lysine methylation states over the entire family of DNA repeats in the mouse genome. Using chromatin immunoprecipitation in a cluster analysis representing repetitive elements, our data demonstrate the selective enrichment of distinct H3-K9, H3-K27 and H4-K20 methylation marks across tandem repeats (e.g. major and minor satellites), DNA transpo- sons, retrotransposons, long interspersed nucleotide elements and short interspersed nucleotide elements. Tandem repeats, but not the other repetitive elements, give rise to double-stranded (ds) RNAs that are further elevated in embryonic stem (ES) cells lacking the H3-K9- specific Suv39h histone methyltransferases. Importantly, although H3-K9 tri- and H4-K20 trimethylation appear stable at the satellite repeats, many of the other repeat- associated repressive marks vary in chromatin of differ- entiated ES cells or of embryonic trophoblasts and fibroblasts. Our data define a profile of repressive histone lysine methylation states for the repetitive complement of four distinct mouse epigenomes and suggest tandem repeats and dsRNA as primary triggers for more stable chromatin imprints. The EMBO Journal (2005) 24, 800���812. doi:10.1038/ sj.emboj.7600545 Published online 27 January 2005 Subject Categories: chromatin & transcription Keywords: DNA repeats double-stranded RNA histone lysine methylation H3-K9, H3-K27 and H4-K20 methylation states mouse epigenome Introduction Over the last years, the genome sequencing of several model organisms revealed that mammals, as compared to unicellu- lar organisms or invertebrates, have a highly complex gen- ome organization, largely resulting from the accumulation of repetitive elements and noncoding sequences (Lander et al, 2001 Waterston et al, 2002). In mouse, such elements account for the majority of its DNA content (44% repetitive and 52% noncoding), whereas only 4% encodes for protein function. As a result, most mammalian genes are disrupted with long intervening sequences that often also contain interspersed repeats. Since the pioneering work of Muller (1930) and McClintock (1951), nonspecific or repetitive sequences have been thought of as ���epigenetic elements��� that can modulate gene expression programmes and also organize heterochromatic domains at centromeres and telomeres (Pardue and Gall, 1970). These repetitive elements range from short interspersed transposa- ble elements to large centromere-associated and telomeric arrays of DNA (Waterston et al, 2002). Although the cluster- ing of repetitive elements in large segments contributes to specialized structures in pericentric heterochromatin, most of the repetitive elements pose an inherent burden to genome stability, as their mobilization facilitates recombination between nonhomologous loci, leading to chromosomal dele- tions and translocations (Kazazian, 2004). Multicellular organisms have developed silencing mechan- isms to prevent remobilization of transposons. These include RNAi-triggered silencing (Ratcliff et al, 1997 Mette et al, 2000 Vastenhouw and Plasterk, 2004), DNA methylation (Jaenisch and Bird, 2003 Bourc���his and Bestor, 2004) and histone modifications (Jenuwein and Allis, 2001). Current data have given rise to models in which transcription across DNA repeats would induce formation of double-stranded RNA (dsRNA), which in turn recruits repressive chromatin modifications and DNA methylation to the underlying chro- matin template (Tamaru and Selker, 2001 Hall et al, 2002 Volpe et al, 2002 Chan et al, 2004 Lippman and Martienssen, 2004 Verdel et al, 2004). Recent mapping studies across large chromosome regions revealed a strong correlation between DNA repeats, noncoding RNA, histone H3 lysine 9 methyla- tion and DNA methylation (Lippman et al, 2004). Similarly, transcription factor mapping along human chromosomes (Cawley et al, 2004) indicated that these factors are not only found at promoters, but also can be identified at non- coding regions. There are three repressive histone lysine methylation marks (H3-K9, H3-K27 and H4-K20) and three distinct methyl- ation states (mono-, di- and trimethylation). H3-K9 trimethyl- ation (Peters et al, 2003) and H4-K20 trimethylation (Schotta et al, 2004) are concentrated at pericentric and centric heterochromatin (Lehnertz et al, 2003). By contrast, H3-K27 trimethylation is enriched at the inactive X chromosome (Plath et al, 2003 Silva et al, 2003 Kohlmaier et al, 2004). Although this chromosome has the highest density of DNA repeats, the contribution of these repeats in X-chromosome inactivation remains unclear. Similarly, H3-K27 trimethyla- tion has been implicated in Polycomb-dependent gene silen- cing (Ringrose and Paro, 2004) via Polycomb response elements (PREs), which also contain short repetitive ele- ments (Ringrose et al, 2003). Despite these parallels, it is Received: 3 November 2004 accepted: 13 December 2004 published online: 27 January 2005 *Corresponding author. Research Institute of Molecular Pathology (IMP), The Vienna Biocenter, Dr Bohrgasse 7, 1030 Vienna, Austria. Tel.: �� 43 1 797 30 474 Fax: �� 43 1 798 7153 E-mail: jenuwein@imp.univie.ac.at 1These authors contributed equally to this work The EMBO Journal (2005) 24, 800���812 | & 2005 European Molecular Biology Organization | All Rights Reserved 0261-4189/05 www.embojournal.org The EMBO Journal VOL 24 | NO 4 | 2005 & 2005 European Molecular Biology Organization EMBO THE E M B O JOURNAL THE EMBO JOURNAL 800

largely unknown whether the same or different histone lysine methylation marks are recruited to large arrays or to inter- spersed repeats and whether these epigenetic states are stably inherited across distinct cell types. Chromatin modifications have been shown to be highly dynamic at heterochromatin during differentiation (O���Neill and Turner, 1995), and recent data indicated significant differences in occupancy of tran- scription factor binding in a genome-wide analysis in Saccharomyces cerevisiae comparing various transcriptional states (Harbison et al, 2004). Here, we analysed the distribution of all nine repressive histone methylation states (H3-K9, H3-K27, H4-K20 mono-, di- and trimethylation) across the repetitive complement of the mouse genome. Using directed and array-based chroma- tin immunoprecipitation (ChIP), we compared tandem satel- lite repeats (pericentric and centric heterochromatin) with distinct families of interspersed repeats including DNA trans- posons, long terminal repeats (LTRs), long interspersed nucleotide elements (LINEs) and short interspersed nucleotide elements (SINEs). We detect distinct chromatin modification patterns between tandem and the various interspersed re- peats, which are further reflected by differences in the non- coding and dsRNAs that are generated from these elements. In addition, the observed repeat-associated histone lysine methylation profiles display significant variability in chroma- tin of different cell types (e.g. embryonic stem (ES) cells, fibroblasts and trophoblasts). Together, our data provide a representative cluster analysis for repressive chromatin mod- ifications of the repetitive part of four distinct mouse epigen- omes. Results Cluster analysis of repetitive elements Mouse chromosomes are acrocentric and contain cytologi- cally visible heterochromatin around their centromeres (Figure 1A). This constitutive heterochromatin can be sub- divided into domains of tandem arrays of A/T-rich major satellite repeats that can comprise X10 000 copies of a 234 base-pair (bp) unit per pericentric region. Similarly, centric heterochromatin (the primary constriction) consists of tan- dem arrays of B2000 copies of 123 bp minor satellite repeats. Both major and minor satellite repeats account for B3.5% of the mammalian genome (Lander et al, 2001 Waterston et al, 2002) (Figure 1B). Interspersed repeats are singular repetitive elements that are integrated over the entire genome. Around 1% of inter- spersed repeats are DNA transposons (e.g. Mariner, Tigger, URR1 and Charlie), which are inactive remnants of a mutated DNA transposase element (Kazazian, 2004). The most abun- dant class of interspersed repeats is represented by different subtypes of retro- or RNA transposons, which together account for 25% of the mouse genome. Subclass I are the LTR transposons, including the highly active intracesternal A particle (IAP) elements. These resemble retroviruses and are associated with over 10% of all spontaneous mutations in mice (Waterston et al, 2002). Subclass II are non-LTR transposons or LINE (or L1) elements, which form the single largest fraction (B19%) of interspersed repeats in the mouse. It has been estimated that between 0.5 and 1% of LINEs are potentially active (Goodier et al, 2001). Subclass III are SINEs, which can be further subdivided into SINE B1, the human counterpart of which are Alu elements, B2 and RSINEs. SINEs are nonautonomous elements and are thought to rely on LINEs for retrotransposition (Smit, 1996). For all of these distinct repeat classes, specific primers for ChIP analyses were designed, which will generate, on aver- age, 200 bp fragments from within each repetitive unit or over the LTR and UTR segments of the various transposons (see Figure 1B). Although this strategy will not allow measure- ment of potential differences in repeat-associated chromatin modifications that may be present at various chromosomes and cannot discriminate solitary copies of interspersed re- peats, it offers a solid cluster analysis of chromatin at DNA repeats, which gauges the sum of a given histone modifica- tion over these distinct repetitive elements. The profile of repressive histone lysine methylation states at distinct repeat classes Lysates of crosslinked chromatin from wild-type (wt) and Suv39h double null (dn) mouse ES cells were sonicated to generate fragments of approximately 300���1500 bp. Chro- matin fragments were analysed by DNA blot to confirm representation of DNA repeats (see Supplementary Figure S1). The sonicated lysates were then used in ChIP with our panel of H3-K9, H3-K27 and H4-K20 antibodies. Precipitated DNA was analysed by real-time PCR with the repeat-specific primer sets (see Figure 1B). As controls, we included primers for ribosomal DNA, actin and tubulin and also performed ChIP with an active mark, histone H3-K4 trimethylation (Santos-Rosa et al, 2002). Based on these control ChIP, we observed that most modifications show a dispersed, basal level signal, which remained below 0.5% of precipitated material (Figure 2, bottom panels). We therefore applied the 0.5% threshold to evaluate accumulation of distinct histone lysine methylation marks across repetitive elements. Using this threshold, we observe selective enrichment for H3-K9 tri-, H3-K27 mono- and H4-K20 trimethylation across mouse major and minor satellite repeats as previously re- ported (Peters et al, 2003 Schotta et al, 2004). This enrich- ment of H3-K9 trimethylation was not detected at human centromeric chromatin (Sullivan and Karpen, 2004). Similar to the apparent under-representation of H3-K9 trimethylation in Drosophila (Ebert et al, 2004) or Arabidopsis thaliana (Jackson et al, 2004) heterochromatin, species-specific altera- tions in satellite composition could account for these differ- ences. For both major and minor satellites, the trimethyl marks are significantly impaired in the control ChIP with chromatin from Suv39h dn ES cells (Figure 2). DNA transpo- sons, exemplified by Mariner and Charlie, are also enriched for H3-K9 trimethylation in a Suv39h-dependent manner. In addition, DNA transposons also contain H4-K20 dimethyla- tion as a second signature mark, although this profile is not consistently associated with different members of DNA trans- posons, such as, for example, Tigger and URR1 (data not shown). For IAP LTRs, we detect H4-K20 trimethylation as the sole prominent mark. Since only a modest reduction in the level of H4-K20 trimethylation is observed in Suv39h dn chromatin, establishment of this mark is probably independent of the Suv39h enzymes and may occur in a mechanism that is distinct from the induction of H4-K20 trimethylation at pericentric heterochromatin (Schotta et al, 2004). For LINEs, hardly any signals are detectable above threshold (e.g. H3-K9 Histone lysine methylation profiling at DNA repeats JHA Martens et al & 2005 European Molecular Biology Organization The EMBO Journal VOL 24 | NO 4 | 2005 801