Transposable elements and their remnants constitute a substantial fraction of eukaryotic genomes. Host genomes have evolved defence mechanisms, including chromatin modifications and RNA interference, to regulate transposable elements. Here we describe a genome surveillance mechanism for retrotransposons by transposase-derived centromeric protein CENP-B homologues of the fission yeast Schizosaccharomyces pombe. CENP-B homologues of S. pombe localize at and recruit histone deacetylases to silence Tf2 retrotransposons. CENP-Bs also repress solo long terminal repeats (LTRs) and LTR-associated genes. Tf2 elements are clustered into ‘Tf’ bodies, the organization of which depends on CENP-Bs that display discrete nuclear structures. Furthermore, CENP-Bs prevent an ‘extinct’ Tf1 retrotransposon from re-entering the host genome by blocking its recombination with extant Tf2, and silence and immobilize a Tf1 integrant that becomes sequestered into Tf bodies. Our results reveal a probable ancient retrotransposon surveillance pathway important for host genome integrity, and highlight potential conflicts between DNA transposons and retrotransposons, major transposable elements believed to have greatly moulded the evolution of genomes.
Transposable elements, prevalent in most eukaryotic genomes, exert diverse effects on their hosts, profoundly influencing the organization, integrity and evolution of the host genome, and the host transcriptome1,2. Host cells have devised strategies, such as DNA and histone methylation and RNA interference (RNAi), to control transposable element activity3,4,5,6. Whereas RNAi has a prominent role in silencing transposable elements in most organisms, in mammals, an adaptive RNAi response, thus far, has been observed only in germ cells7. Moreover, RNAi in S. pombe is known to target preferentially a specific class of repeat (dg and dh) elements associated with constitutive heterochromatic regions, but RNAi and heterochromatin machineries have minor roles in silencing Tf2 retrotransposons and solo LTRs8 dispersed throughout the genome9,10. Therefore, additional RNAi-independent mechanisms probably exist to recognize and silence transposable elements.
The long evolutionary presence of transposable elements in eukaryotic genomes has resulted in many instances of hosts ‘domesticating’ transposable-element-encoded factors to perform cellular functions11,12,13. Human CENP-B protein, which facilitates centromere formation by binding to short repeats within centromeric alpha satellite DNA14,15, is derived from transposases of pogo DNA transposons16,17. CENP-B, highly conserved in mammals, has homologues in other systems13,18. The genome of S. pombe encodes three CENP-B homologues: Abp1, Cbh1 and Cbh2 (ref. 19). These have been shown to have redundant roles in centromeric heterochromatin formation and chromosome segregation19,20. However, unlike cells lacking heterochromatin proteins such as Swi6/HP1, loss of Abp1 results in a slow-growth phenotype exacerbated by additional deletions of other CENP-B homologues18,19, suggesting additional functions by CENP-Bs.
Here, we report roles for CENP-B homologues in the surveillance for retrotransposons and in genome organization in S. pombe. CENP-Bs bind to Tf2 retrotransposons and their remnants, and mediate Tf2 silencing by recruiting the class I histone deacetylase (HDAC) Clr6 (refs 21, 22) and the class II HDAC Clr3 (ref. 23). CENP-Bs also repress several genes through nearby LTRs and facilitate HDAC recruitment to heterochromatic loci. We demonstrate that Tf2 retrotransposons scattered across the genome are clustered into Tf bodies, the organization of which is mediated by CENP-Bs. This study highlights a potentially hitherto unrecognized host genome surveillance mechanism that exploits the targeting of a specific class of transposable elements by proteins of another class to rein in transposable element activities.
Genome-wide distributions of Abp1 and Cbh1
To gain insights into disparate phenotypes exhibited by mutations in the CENP-B family members of S. pombe, we mapped the distributions of these factors across the genome. Chromatin immunoprecipitation coupled with DNA microarray (ChIP–chip) analyses revealed Abp1 and Cbh1 binding at numerous loci on all three chromosomes (Fig. 1a). Consistent with the original finding of Abp1 association with certain ARS sites24, we found Abp1 enrichment at ars3002 (Supplementary Fig. 1a). Abp1 was also detected at centromeres (Supplementary Fig. 1b)19,20. Close examination revealed preferential Abp1 enrichment at sequences immediately outside of dh repeats of centromere I and II, although low levels of binding were also observed at other sites (Supplementary Fig. 1b). Notably, Abp1 and Cbh1 were highly enriched at Tf2 retrotransposons (Fig. 1a). Abp1 and Cbh1 co-localized throughout single, tandem and partial Tf2 elements with both proteins displaying distinct binding peaks at the 5′ and 3′ LTRs (Fig. 1b and Supplementary Fig. 1c, d), suggesting a possible mode for recruiting these proteins to Tf2 via specific Abp1/Cbh1-binding sites within Tf2 LTRs. Conventional ChIP with primers positioned at unique sequences confirmed binding of Abp1 and Cbh1 at all 13 full-length Tf2 elements and Tf2 fragment 1 (Fig. 1b and Supplementary Figs 1c, d and 2). We also found that Abp1 and Cbh1 localized to solo LTRs, wtf elements (repeats often associated with Tf LTRs) and intergenic regions, many of which associate with the promoters of nearby genes (Fig. 1a, c and Supplementary Figs 3 and 4).
CENP-Bs silence Tf2 and LTR-associated genes
The binding of CENP-Bs to distinct genomic sites prompted us to investigate whether loss of these factors influences target loci expression. Tf2 expression was monitored at two sites: LTR, which shares high sequence similarity with some solo Tf2 LTRs, and coding region, which is present only in full-length Tf2 and Tf2 fragment 1 elements. Assays at both sites showed marked increases in Tf2 expression in an abp1Δ strain (Fig. 2a). However, loss of cbh1 or cbh2, or both, did not lead to detectable increases in Tf2 expression; yet, single or double deletions of these genes in the abp1Δ background contributed to a slight further increase in Tf2 expression relative to the abp1Δ mutant strain alone. We extended expression analyses to wtf elements and several genes close to solo LTRs enriched for CENP-Bs, and found that similar to Tf2, loss of Abp1 accounted for most of the increased levels of expression observed at wtf elements and genes associated with LTRs (Fig. 2b, c and Supplementary Fig. 4). These analyses suggest that among the S. pombe CENP-Bs, Abp1 is the member critically required for preventing widespread expression of Tf2, solo LTRs and LTR-associated genes, and thus could explain the marked growth defect observed in abp1Δ cells18.
Although both Abp1 and Cbh1 co-localize at Tf2, Abp1 alone is sufficient to repress Tf2 expression, suggesting that Abp1 could be responsible for recruiting Cbh1 to Tf2. Indeed, the absence of either Cbh1 or Cbh2, or both, had no discernible effect on Abp1 binding to Tf2 (Fig. 2d), whereas Cbh1, although unaffected by the absence of Cbh2, was delocalized from Tf2 in cells lacking Abp1 (Fig. 2e). Immunoprecipitation assays showed Abp1 in complex(es) with Cbh1 and Cbh2 (Fig. 2f), supporting the idea that Abp1 recruits Cbh1, and possibly Cbh2, to Tf2.
Abp1 recruits HDACs to Tf2 and a heterochromatic locus
The HDACs Clr3 (a component of the SHREC (Snf2/Hdac-containing repressor complex) silencing complex23) and Clr6 (which exists in multi-subunit complexes21) have been implicated in controlling Tf2 expression in S. pombe10,21,23. We investigated potential genetic interactions between HDACs and Abp1 in Tf2 silencing. Tf2 expression was increased and further elevated in clr6 and double mutant clr3 clr6 cells, respectively (Fig. 3a)10. However, cumulative increase in Tf2 expression in double HDAC mutant cells was less compared to that observed in abp1Δ cells or in cells deficient for Abp1 and either Clr3 or Clr6 (Fig. 3a). These data indicate that Abp1 might act upstream by recruiting these HDACs to silence Tf2. Indeed, Abp1 enrichment at Tf2 was not affected in single or double HDAC mutants, whereas Clr3 and Clr6 localizations at Tf2 were compromised in abp1Δ cells (Fig. 3b, e). Also, Cbh1 localization at Tf2 was partially affected in HDAC mutant cells (Fig. 3b), suggesting that HDACs might aid in the recruitment of Cbh1 to LTRs. We found Abp1 interacting with Clr3 and Clr6 in vivo (Fig. 3c, d). Together, these data show that Abp1 negatively regulates Tf2 expression, in part, by directly recruiting HDACs to Tf2.
We next determined whether Abp1-mediated silencing of retrotransposons intersects with that of HIRA/Hip1 histone chaperone, implicated in Tf2 silencing25. The levels of Tf2 expression were higher in abp1Δ hip1Δ cells compared with single mutant abp1Δ or hip1Δ cells (Fig. 3f), indicating that Abp1 and Hip1 might silence Tf2 through distinct pathways. In this regard, Abp1 and Cbh1 localizations at Tf2 were not affected in hip1Δ cells (Fig. 3g).
Heterochromatic silencing also involves HDAC recruitment to the target loci26. Previous analysis revealed three major SHREC peaks at the mating-type (mat) region23, two of which appear to correspond to silencer elements near the silent mat cassettes (Fig. 4a). The third peak overlaps with CENP-B binding outside of cenH (Fig. 4a), a known RNAi-dependent heterochromatin nucleation centre26, suggesting that in addition to Tf2, CENP-Bs might also recruit SHREC at mat. Indeed, Clr3 binding was impaired in abp1Δ cells specifically at CENP-B binding sites (Fig. 4b). However, considerable levels of Clr3 binding remain across the silent mat region in abp1Δ cells, probably due to redundant recruitment mechanisms26.
Dispersed Tf2 elements cluster into Tf bodies
To gain further insights into the functions of CENP-Bs, we explored their subcellular localization. Immunostaining analyses showed that all three CENP-B proteins display complex nuclear ‘network’ structures (Fig. 5a and Supplementary Fig. 5a). In addition to localizing to 4,6-diamidino-2-phenylindole (DAPI)-stained nuclear regions, a dense concentration of CENP-Bs could be seen near the nucleolus. The congression of CENP-Bs into nuclear networks prompted us to investigate whether Tf2 elements are organized into higher-order structures. We tested this by performing fluorescence in situ hybridization (FISH) using probes directed against the coding region of Tf2. Notably, despite the presence of 13 full-length Tf2 elements in the genome, only 1–3 Tf2 foci could be seen per cell nucleus in wild-type cells (Fig. 5b), suggesting that CENP-Bs may have a role in clustering Tf2 into specialized structures that we refer to as Tf bodies. Indeed, a substantial fraction of single and double mutant CENP-B cells contained more than three Tf2 spots (Fig. 5b, c).
Mammalian CENP-Bs form homodimers through their conserved carboxy-terminal domain27,28. We explored whether Abp1 could form dimers that might contribute to ‘bundling’ Tf2 elements into Tf bodies. We found that Abp1 interacts with itself (Supplementary Fig. 6a), and this interaction depends on its dimerization domain. Unlike abp1Δ, which affects genomic stability that could potentially contribute to Tf2 de-clustering, truncation of the Abp1 dimerization domain (abp1-dm) did not result in gross defects in growth or chromosome segregation (Supplementary Fig. 6b). Also, abp1-dm cells showed little or no change in Tf2 expression (Supplementary Fig. 6c). However, abp1-dm cells were slightly defective in Tf2 clustering, a phenotype that was further exacerbated when cbh1 and cbh2 deletions were combined with abp1-dm (Fig. 5d).
De-clustering of Tf2 elements upon oxidative stress
Host cells subject to stress can activate proviruses and silent transposons29,30. Tf2 expression is known to be upregulated when cells are exposed to oxidative stress31. We assessed the effect of oxidative stress on the integrity of Tf bodies. Transient exposure of cells to hydrogen peroxide caused de-clustering of Tf2 elements, despite CENP-Bs still being bound to Tf2 (Supplementary Fig. 7a, b). It is possible that signalling pathways required for the cellular response to environmental stresses modify Tf bodies to facilitate rapid restructuring of the genome.
CENP-Bs target an extinct retrotransposon
In addition to Tf2, the S. pombe genome also contains 249 solo LTRs and LTR fragments, of which 11% belong to an extinct retrotransposon, Tf1 (refs 8, 32). Because we found high levels of Abp1 and Cbh1 at Tf1 LTRs, we hypothesized that CENP-Bs could also target Tf1. We tested this idea by reintroducing a full-length copy of Tf1 into the genome. Previous work has shown that Tf1 can insert into the genome either via integrase-mediated transposition or homologous recombination33. Expression of Tf1-neo from a plasmid showed a minor increase in Tf1 transposition in CENP-B mutant cells compared to wild-type cells (Supplementary Fig. 8a). However, we observed a marked increase in frequencies of genomic insertion by Tf1 in cells lacking CENP-Bs, in particular abp1Δ or cbh1Δ mutants, when an integrase-defective Tf1 was used (Supplementary Fig. 8a). This result indicated that Abp1 and Cbh1 could block the homologous recombination of Tf1. Sequencing of several Tf1 integrants from abp1Δ and cbh1Δ cells showed Tf1 recombination predominantly with existing Tf2 elements or their remnants (Supplementary Fig. 8b). Therefore, in addition to silencing, CENP-Bs suppress recombination at Tf elements, which has important implications for maintenance of genomic integrity.
We next investigated whether the CENP-B-based surveillance mechanism can recognize a Tf1 element integrated into the genome. Insertion of Tf1 at SPAC7D4.08, a euchromatic gene, resulted in the targeting of CENP-Bs to this locus (Fig. 6a). CENP-Bs suppress transcription and transposition of the Tf1 integrant, as indicated by increases in expression and transposition frequency of Tf1 in the abp1Δ background (Fig. 6b, c). We also found that insertion of Tf1 alters the localization of the SPAC7D4.08 locus in the nucleus. Whereas SPAC7D4.08 does not usually associate with Tf bodies, Tf1 insertion at SPAC7D4.08 resulted in a high incidence of SPAC7D4.08 co-localizing with Tf bodies (Fig. 6d and Supplementary Fig. 5b), probably as a result of Tf1 being ‘bundled’ into Tf bodies.
Our study uncovers unexpected targeting of S. pombe CENP-B homologues to Tf retrotransposons and their remnants to mediate both transcriptional and recombinational repression. CENP-Bs thought to have originated from transposases encoded by an ancient pogo-like DNA transposon16,17 could hamper the mobility of LTR retrotransposons (this study). Given that both DNA transposons and LTR retrotransposons are flanked by repetitive DNA structures, and transposases bind to terminal inverted repeats of DNA transposons2, it is possible that during evolution a CENP-B precursor acquired the ability to target retrotransposon LTRs, and subsequently was co-opted by the host into its gene repertoire for controlling transposable elements.
CENP-B-mediated Tf silencing is, in part, dependent on CENP-Bs recruiting Clr3-containing SHREC and Clr6 HDAC complexes, which are also required for heterochromatic silencing of centromeric repeats10,21,23. Thus, CENP-B localization at heterochromatic regions could aid in silencing by means of HDAC recruitment. Indeed, Abp1 mediates SHREC recruitment to the silent mat region. The targeting of HDACs via CENP-B might facilitate assembly of ‘closed’ chromatin that not only represses transcription but also protects genomic integrity by rendering repetitive sequences recombinationally inert23,26,34.
Considering that heterochromatin and CENP-Bs recruit a common set of silencing factors, CENP-B regulation of Tf elements might represent a ‘simple’ form of local heterochromatic silencing. In this respect, it would not be surprising that Tf2 silencing might use other heterochromatin and RNAi components whose normally minor or redundant roles manifest under specific conditions. S. pombe CENP-Bs may recruit distinct effectors. Apart from the recruitment of HDACs to Tf elements, CENP-Bs might target additional factors critical for transposable element surveillance. As Abp1 and Cbh1 can bind to distinct loci, and that these factors do not always co-localize with SHREC23, CENP-B binding in contexts other than heterochromatin and transposable elements might recruit activities important for other chromatin transactions, such as transcription and DNA replication.
Accumulating evidence implicates transposable-element-derived sequences as regulators of gene expression in diverse species1,35,36,37. A substantial fraction of human promoters contains transposable element sequences38, and instances of their contribution to gene regulation have been documented1,39. We provide evidence that CENP-Bs bound to LTRs can regulate expression of nearby genes, presumably owing to their ability to recruit chromatin modifiers. CENP-Bs also occupy several gene promoters, that, at one time, might have contained transposable element sequences, but most of which have decayed beyond recognition. Therefore, in addition to retrotransposon surveillance, CENP-Bs and their targeted transposable element sequences might serve as versatile regulatory modules, enhancing the ability of cells to modulate gene regulatory networks35.
Our analyses revealed unexpected clustering of Tf2 elements into Tf bodies. These observations are reminiscent of Drosophila gypsy retrotransposons brought together into specialized bodies to facilitate chromatin organization40. Moreover, transposon-derived MAR/SAR sequences38 have been shown to create chromatin loops41. These findings indicate a conserved role for transposable elements in genome organization. Dimerization of CENP-B proteins42 (this study) bound to LTRs may directly or indirectly promote either local or long-range loops between LTRs of individual Tfs or different Tf elements. As such, enrichment of CENP-B proteins at Tf2 peaks at flanking LTRs but progressively decreases towards the centre of Tf2 elements. The clustering of Tfs into distinct bodies may facilitate transposable element surveillance or other genome-wide processes including prevention of aberrant recombination, coordinated transcriptional control and genome reorganization in response to environmental stresses (Fig. 4g)29.
This study may have implications for several phenomena observed in other systems. Drosophila P and 1360 elements can trigger heterochromatic silencing that is highly dependent on element copy number43,44. This mass action requirement for silencing might reflect a high concentration of specific transposable-element-binding proteins similar to CENP-Bs capable of recruiting chromatin modifiers. The formation of Tf bodies could have parallels in mammalian X-chromosome inactivation, in which transposable elements might facilitate the assembly of a silent nuclear compartment45,46,47. Recent evidence suggests that whereas RNAi has a prominent role in silencing transposable elements in germ cells in higher eukaryotes7, there are probably alternative mechanisms, possibly similar to that of S. pombe CENP-B-based surveillance, for regulating transposable elements in somatic cells.
Strains were generated by standard yeast methods. ChIP and ChIP–chip were performed as previously described9. Quantitative PCR (qPCR) expression analysis was done using a two-step real-time RT–PCR. Mobility of a Tf1 integrant was assessed using Tf1-neo strains carrying an artificial intron (AI)48,49 inserted in the opposite orientation of neo, thereby inactivating neo. Retrotransposition rate was determined by fluctuation analysis50 of cells regaining neo owing to AI loss during Tf1 retrotransposition. Sequences of primers used are available in Supplementary Table 1.
S. pombe strains
abp1Δ, hip1Δ and C-terminal-tagged strains (Abp1–Flag(3×), Cbh1–Flag(3×), Abp1–Myc(13×), Cbh1–Myc(13×), Cbh2–Myc(13×)) were constructed according to a PCR-based method. The Tf1-neoAI strain was generated by transforming a strain containing Tf1-neo integrated at SPAC7D4.08 with a PCR neo fragment containing an artificial intron (AI)48,49 positioned in reverse orientation to neo; neo-sensitive colonies were selected, and AI insertion into neo was confirmed by DNA sequencing. abp1Δ::leu2, cbh1Δ::leu2 and cbh2Δ::leu2 strains were gifts from Y. Murakami. All other strains were obtained by standard genetic crosses.
ChIP and ChIP–chip experiments were performed as previously described9 with minor modifications. Sonicated cross-linked chromatin (50 ml S. pombe culture (optical density (OD) 0.5–1) cross-linked with 3% paraformaldehyde for 30 min at 18 °C followed by 10 mM dimethyl adipimidate for 45 min at room temperature) was immunoprecipitated with anti-Myc (9E10, BD Biosciences; A-14, Santa Cruz Biotech) antibody for Clr3–Myc, or anti-Flag-coupled beads (M2, Sigma-Aldrich) for Abp1–Flag and Cbh1–Flag. Antibody used against Clr6 has been described previously21. For conventional ChIP assay, ChIP PCR products (26–30 cycles) were analysed by gel electrophoresis (4% polyacrylamide gel) and visualized with SYBR Green (Invitrogen) staining. Relative ChIP enrichment was quantified using a Typhoon scanner and ImageQuant software.
Quantitative real-time PCR (qPCR)
A total of 500 ng of DNase-treated total RNA isolated by MasterPure yeast RNA purification kit (Epicentre) was reverse transcribed (RT) with Superscript III and oligo(dT)20 primer (Invitrogen). Complementary DNA (1/50 RT reaction) was used as templates for SYBR green (0.1× dilution) qPCR analysis (Opticon 2). Relative fold expressions relative to act1 were determined in Excel.
A total of 50 ml S. pombe culture (OD 1–1.5) was lysed in HCS buffer (150 mM HEPES pH 7.2, 250 mM NaCl, 0.1% NP-40, 20 mM (NaF, BGP), 1 mM (EDTA, dithiothreitol, PMSF) and protein inhibitor tablet (Roche)) by bead beater (three times 30 s with 2 min interval on ice). Protein extracts (1–5 mg) from indicated strains pre-cleared in protein A/G beads were incubated for 2 h at 4 °C with M2 agarose beads (Sigma-Aldrich), extensively washed afterwards with HCS buffer and subjected to PAGE gel (NUPAGE Novex 10% BT, Invitrogen) and western blot analyses (iBlot, Invitrogen) with either anti-Myc (NB600-336, Novus Biologicals), anti-Flag (M2, Stratagene) or anti-Clr6 antibodies.
For immunofluorescent analysis, 10 ml S. pombe cells (OD 0.5–1) were cross-linked in 1.2 M sorbitol with 3.8% paraformaldehyde at 18 °C for 30 min, quenched with 125 mM glycine, and subjected to cell-wall digestion with zymolase (Seikagaku), and blocked with PEMBAL (100 mM PIPES pH 6.9, 1 mM EGTA, 1 mM MgSO4, 1% BSA, 0.1 M l-lysine) for 1 h before incubation with antibody. Anti-Myc (9E10, BD Biosciences) antibody and Alexa-Fluor-488-conjugated goat anti-mouse antibody (Invitrogen) were used to detect Myc-tagged CENP-B. For FISH analysis, cells were treated similarly as those for immunofluorescent analysis. After blocking with PEMBAL, cells were treated with RNase A (0.1 μg μl-1) at 37 °C for at least 3 h. Hybridization was carried out with 100–150 ng of probes in 100 μl hybridization buffer (50% formamide, 2× SSC, 5× Denhart’s solution, 10% dextran sulphate) at 40 °C for 10–12 h. Cells were washed three times in 100 μl 2× SSC for 30 min each before mounting. For FISH probe preparation, a 3.0-kb or 4.3-kb PCR fragment corresponding to the coding region of a full-length Tf2 element or the upstream promoter region of SPAC7D4.08, respectively, was generated by PCR. Tf2 PCR fragment was subsequently cloned into a TOPO II vector (TOPO II Tf2-orf). Tf2-orf plasmid DNA or purified SPAC7D4.08 PCR product was digested with AluI and DdeI restriction enzymes and labelled by random priming (Takara 6045) with either dCTP-cy3 (Tf2) or amino-allyl dUTP/Alexa 488 (SPAC7D4.08). Statistical significance of Tf2 de-clustering in abp1 mutant cells relative to those of wild-type (P < 0.05) was assessed using a one-way analysis of variance test. For image deconvolution, immunofluorescent analysis and FISH images were collected at 0.2 μm intervals along the Z-axis and subjected to volume deconvolution using the nearest neighbour method (Improvision, OPENLAB).
Wild-type and CENP-B mutants were assayed for Tf1 integration as described previously49. For mobility assay of a Tf1 integrant, 2 × 105 cells of wild-type or CENP-BΔ strains carrying SPAC7D4.08::Tf1-neoAI were inoculated in ten parallel 2-ml cultures and grown at 30 °C for 2–3 days. 1 × 108 cells from each culture were plated on a G418 plate (150 μg ml-1), and neo-resistant colonies were scored. Cells from five cultures were titred on YEA plates. Transposition rate was determined by fluctuation analysis50.
We thank members of the Grewal laboratory for discussions and Y. Murakami for strains. This research was supported by the Intramural Research Program of the National Institutes of Health, National Cancer Institute.
Author Contributions H.P.C. and S.I.S.G. designed research and analysed data. H.P.C. performed most experiments. K.N. contributed strains and ChIP analysis. H.E. and H.L.L. performed the Tf1 reintroduction assay. H.P.C. and S.I.S.G. wrote and edited the paper.
This file contains Supplementary Figures 1-8 with Legends and Supplementary Table 1. Supplementary Figures 1-8 show results from ChIP-chip, conventional ChIP, qPCR, IF and FISH analyses that complement data present in primary Figures 1-6. Supplementary Table 1 contains sequences for all oligonucleotides used in this study.