|Home | About | Journals | Submit | Contact Us | Français|
Shelterin binds and protects mammalian telomeres. Here, we generated cells and mice conditionally deleted for the shelterin component RAP1. We find that Rap1 deficiency is dispensable for telomere capping but leads to increased telomere recombination and fragility. Mice with Rap1 deletion in stratified epithelia are viable but have shorter telomeres and develop skin hyperpigmentation at aduldhood. By performing chromatin immunoprecipitation coupled with ultra-highthroughput sequencing, we find that RAP1 binds to telomeres and to extra-telomeric sites through the (TTAGGG)2 consensus motif. Extra-telomeric RAP1 binding sites are enriched at subtelomeric regions, in agreement with preferential deregulation of subtelomeric genes in Rap1-deficient cells. More than 70% of extra-telomeric RAP1 binding sites are at the vicinity of genes and 31% of the genes deregulated in Rap1-null cells contain RAP1 binding sites, suggesting a role of RAP1 in transcriptional control. These findings place a shelterin component at the interface between telomere function and transcriptional regulation.
RAP1 (or Terf2ip) is part of the shelterin complex at mammalian telomeres, also encompasing TRF1, TRF2, POT1, TPP1 and TIN2 1. Human RAP1, hRAP1, was identified as a TRF2-interacting protein with homology to budding yeast scRap1 2, 3. scRap1 binding to telomeres controls telomere length and subtelomeric silencing 4-10. In addition, scRap1 acts as a transcription factor by controlling the expression of glycolytic enzymes and ribosomal genes 11, 12. While scRap1 binds telomeric DNA directly, hRAP1 its proposed to be recruited to telomeres by TRF2 3. hRap1 over-expressing cells show continuous telomere lengthening 3. Potential roles of mammalian RAP1 in transcriptional regulation and subtelomeric gene silencing are unknown.
Here, we study the role of mouse RAP1 in telomere biology and its impact on gene expression programs and organismal viability by generating Rap1 tissue-specific conditionally deleted cells and mice. We show that mammalian RAP1 is dispensable for telomere capping but prevents telomere recombination and fragility. Similar findings have been recently reported for an independent conditional knock-out allele of Rap1 13. In contrast to analogous mouse models for TRF1 and TPP1 deficiency 14, 15, targeted Rap1 deletion to stratified epithelia does not impact on mouse viability but leads to skin hyperpigmentation at adulhood and presence of shorter telomeres.
Besides its telomeric roles, we demonstrate a role of RAP1 in repression of subtelomeric genes, analogous to that of yeast andTrypanosoma brucei RAP1 proteins 16. Moreover, we find that RAP1 binds to both telomeric and extra-telomeric TTAGGG repeats. Extra-telomeric RAP1 binding sites are enriched at subtelomeric regions, in agreement with preferential derepression of subtelomeric genes in Rap1-deficient cells. Notably, more than 70% of RAP1 binding sites are found at intragenic positions or at the vicinity of gene-coding chromatin. By performing gene expression studies, we further show that more than a third of the genes significantly deregulated in Rap1-deficient cells contain RAP1 binding peaks. In summary, we identify RAP1 as an important factor for telomere integrity and for transcriptional gene regulation in mammals. These findings provide evidence that a shelterin component can bind to non-telomeric regions and have extra-telomeric functions.
We generated a conditional Rap1 knockout mouse, Rap1flox/flox, by flanking Rap1 exon 3 (E3) by loxP sites (Fig. 1A). We included frt sites to remove the neomycin gene (Fig. 1A) 17. This targeting strategy does not to interfere with the essential gene Kars 18, which shares promoter with Rap1. Deletion of Rap1 E3 eliminates the Rap1 telomere localization domain (RCT) and the nuclear localization signal (NLS)(Fig. 1B). Deletion of Rap1 E3 also disrupts of the 3′ UTR and polyadenilation signal.
Wild-type and Rap1flox/flox mouse embryonic fibroblasts (MEFs) infected with a pBabe-Cre retrovirus showed poor proliferation in vitro (Fig. S1A). Both genotypes proliferated normally by downregulating p53 using a small hairpin RNA (shRNA) against p53 (Rap1Δ/Δ-shp53-Cre; Fig. 1C and Fig. S1B) or by cancelling both the p53 and Rb pathways using SV40 large T (LT) antigen (Rap1Δ/Δ-LT-Cre; Fig. 1C and Fig. S1C). Rap1Δ/Δ-LT-Cre MEFs also showed a similar cell cycle profile to control MEFs (Fig. S1D).
Rap1Δ/Δ LT-Cre cells showed normal expression of Rap1 exons 1 and 2 and of Kars mRNA, while Rap1 E3 transcripts were undetectable (Fig. 1D). Two antibodies against full-length RAP1 recognized a low abundance protein (25-fold less abundant than full-length RAP1) of approximatelly 40 kD in Rap1Δ/Δ-LT-Cre cells that could correspond to a truncated RAP1 protein (RAP1Δ, Fig. 1E). RAP1 antibodies immunoprecipitated RAP1Δ only in Rap1-null MEFs but not in wild-type controls (Fig. 1F). RAP1Δ was undetectable in nuclear extracts (Fig. 1E) or in the chromatin-bound fractions (Fig. 2A) of Rap1Δ/Δ-LT-Cre cells, in agreement with deletion of the NLS and RCT regions. We confirmed defective RAP1 binding to telomeres in Rap1-null cells by immunofluorescence (Fig. 1G; Fig. 2E) and ChIP-sequencing experiments (Fig. 7A). Based on the low and restricted (cytoplasm) expression of RAP1Δ, it is unlikely that it might exert significant dominant negative effects at telomeres in Rap1-null cells, although potential dominant negative effects on other cellular pathways cannot be discarted.
In chromatin fractionation assays, binding of TRF1, TRF2, POT1, TPP1 and TIN2 to chromatin was normal in the absence of RAP1 protein (Fig. 2A,B). As control, histone H3 showed normal binding to chromatin in both genotypes (Fig. 2A). These results were extended to telomeric chromatin by using chromatin immunoprecipitation (ChIP) assays (Fig. 2C,D), as well as by immunofluorescence staining, which showed normal binding of TRF1, TIN2, TRF2 and TPP1 to RAP1-depleted telomeres (Fig. 2E).
To address DNA damage at telomeres in the absence of RAP1, we determined their co-localization with γH2AX foci, the so-called TIFs 19, 20. Rap1 deletion 21 resulted in > 70% of Rap1Δ/Δ cells showing > 3 TIFs per metaphase, while these events were rare in control cells (Fig. 3A-D). We detected phosphorylation of the CHK1 and CHK2 checkpoint kinases in Rap1Δ/Δ cells but not in wild-type cells (Fig. 3E). As control, TRF1-null cells showed robust activation of these kinases 14 (Fig. 3F). Chemical inhibition of CHK1 and CHK2 rescued both cells with DNA damage and cells showing > 3 TIFs (Fig. 3G-I), indicating that both pathways are activated upon Rap1 deletion.
Rap1 deletion did not lead to telomere fusions in MEFs (Fig. 3J). Instead, Rap1-deleted MEFs showed increased frequencies of chromosome ends with more than one telomere signal or multitelomeric signals (MTS; Fig. 3J), recently related to telomere fragility 14, 22-24. MTS were increased in wild-type MEFs treated with drugs that induce telomere fragility, such as aphidicolin, and further increased in similarly treated Rap1-deficient MEFs (Fig. 3J). Chromosome breaks were also increased in aphidicolin-treated Rap1-null MEFs compared to wild-type controls (Fig. 3J). These results indicate that RAP1-depleted telomeres are prone to fragility but not fusion.
Sister telomere recombination (T-SCE) frequencies were also elevated in Rap1Δ/Δ-Cre-LT MEFs compared to wild-type controls (Fig. 3K,L), supporting the recently proposed role of RAP1 in preventing telomere recombination 13.
Finally, telomere quantitative FISH (Q-FISH) analysis on metaphases, indicated slightly shorter telomeres in passage 3 Rap1Δ/Δ-Cre-LT MEFs compared to wild-type controls, including increased abundance of short telomeres and decreased abundance of long telomeres (Fig. S2A-C). This shortening was not associated to decreased TERT mRNA and protein levels, or to lower in vitro telomerase activity in Rap1-null MEFs (Fig. S2D-F). A number of recently identified TERT target genes 25, also showed a similar expression in wild-type and Rap1-null MEFs (Fig. S2F).
Next, we targeted Rap1 deletion to stratified epithelia in Rap1Δ/ΔK5-Cre mice (Fig. 4A) 26, 27. These mice show excision of Rap1 E3 in the epidermis (E) but not in the dermis (D) (Fig. 4A) in agreement with K5-Cre expression pattern 27. Accordingly, RAP1 immunofluorescence was undetectable in Rap1Δ/ΔK5-Cre epidermis (Fig. 4B,C).
Rap1Δ/ΔK5-Cre mice are viable and survive to adulthood (Fig 4D), however, they develop skin hyperpigmentation at aduldhood (Fig. 4E). In line with this, Rap1Δ/ΔK5-Cre epidermis showed shorter telomeres (Fig. 4F,G) and increased abundance of cells with DNA damage (γH2AX-positive) (Fig. 4H) compared to age-matched controls.
To address a role for RAP1 in subtelomeric gene regulation, we performed gene expresion analysis of wild-type and Rap1-null MEFs. We pre-defined subtelomeric (S) regions as the 3Mbp regions adjacent to telomeres, which allowed us to recruit a robust number of genes located in S regions (n=917) and a total of 40260 genes located in non-subtelomeric (NS) regions. Genes located in S regions showed significant differences in trancriptional activity when compared to genes located in NS regions (p= 1.044×10−5, Fig. 5A). In both gene sets (S and NS), the median distribution of log2 Rap1+/+/Rap1Δ/Δ values was displaced from zero (Fig. 5A), indicating a genome-wide effect of Rap1-deletion on gene expression. This effect was more pronounced in S regions compared to NS regions (median displacement 0.04 versus 0.01, respectively; Fig. 5A). Moreover, S regions in Rap1-deficient cells were significantly overexpressed compared to S regions in wild-type cells (p<10−6; Fig. 5A).
Gene set enrichment analysis (GSEA) showed that S regions in chromosomes 1, 2, 11, 15 and 19 were significantly enriched (FDR<0.25) in Rap1-deficient cells compared to wild-type cells (Fig. 5B), while only one S region in chromosome 7 showed the inverse pattern (Fig. 5C). This is also ilustrated by representing the log2 Rap1Δ/Δ/Rap1+/+ expression value for each individual S gene at its chromosomal location (Fig. 5D). Thus, Rap1 deletion results in gene expression changes that are significantly more pronounced at subtelomeric regions, where it primarely leads to augmented gene expression. Finally, this effect was not associated with changes in subtelomeric DNA methylation 28 (Fig. S3A,B), global DNA methylation (B1 SINE repeated elements; Fig. S3C), or telomeric transcription (TERRA, Fig. S3D,E) 29, 30.
Apart from changes in subtelomeric gene expression, a stringent (FDR<0.15) differential gene expression analysis revealed 15 genes significantly downregulated and 1 gene significantly upregulated in Rap1-null MEFs (FDR<0.15)(Fig. 6A; Tables S1 and S2). All 16 differentially expressed genes (DEGs) were validaded by quantitative PCR (q-PCR)(Fig. 6B). Rap1 was the most dowregulated gene (>8-fold by microarray analysis and > 200-fold by qPCR analysis) (Fig. 6A,B; Table S1), while the rest of DEGs changed by >4-fold (−2>logFC>2; Table S1). Interestingly, 8 out of the 16 DEGs are related to cancer as indicated by GSEA (Fig. 6A-C).
GSEA and Gene Ontology (GO) analysis on genes dowregulated by at least 2-fold in Rap1-null MEFs (logFC<-1) extended the analysis to a total of 478 genes (Tables S3 and S4-6), including Igf2 (Fig 6B). The downregulated genes H19 and Igf2 (Fig. 6A,B) are imprinted genes. GSEA for imprinted genes confirmed transcriptional repression of many imprinted genes in Rap1-null cells (FDR<0.01)(Fig. S4A; Tables S3 and S7). GSEA analysis also revealed significant downregulation of genes involved in cell adhesion and metabolism in Rap1-null cells (Fig. S4B-F; Tables S8-11). These included genes of the insulin secretion, PPAR signaling, and growth hormone (GH) pathways, including Igf2 (Fig. 6B). In turn, Rap1-null cells showed significant upregulation of ABC transporters (FDR=0.216) and genes involved in Type II Diabetes (FDR=0.067; Fig. S4D; Table S12), suggesting a negative impact of Rap1 deletion on metabolism.
Two highly downregulated genes in Rap1-null cells, NNMT (Nicotinamide N-methyltransferase), a regulator of lifespan linked to the sirtuin pathway, and CTGF (conective tissue growth factor) (Fig. 6A,B), were recently found upregulated in calorie restricted (CR) mice 31. Comparison of the CR gen set 31 to genes downregulated in Rap1-null MEFs indicated that CR gene set is significantly downregulated in Rap1-null cells (Table S13), with both gene sets showing an inverse pattern (Fig. 6D). This includes PGC1α, a regulator of lipid metabolism 32, which we further confirmed by qPCR (5-fold decrease in Rap1-null MEFs compared to controls, Fig. 6E). Together, these results indicate that absence of Rap1 leads to gene expression changes with a predicted negative impact on metabolism.
We noted that Rap1Δ/ΔK5-Cre female mice weighted approximatelly 10% more than age-matched wild-type littermates from week 14 of age onwards, a phenotype which was not apparent at birth and was not observed in males (Fig. 4D; Fig. S5A,B). To address whether this could be related to abnormal expresion of key metabolic genes upon Rap1 abrogation, we subjected male and female mice to either a standard diet (SD) or a high fat diet (HFD) (Fig. S5C,D). While wild-type and Rap1Δ/ΔK5-Cre male mice gained weight following placement on HFD compared to the same genotypes in SD, Rap1Δ/ΔK5-Cre females on a SD gained weight at the same rate than wild-type females on a HFD under conditions of similar food intake (Fig. S6C-E), and this was not increased by placing them on a HFD. Thus, Rap1 deletion in females phenocopies exposure to a HFD. A similar phenotype of increased body weight in female mice is reported for PGC1α-deficiency 33.
To study the role of RAP1 in subtelomeric gene silencing and transcriptional regulation, we set to determine the in vivo RAP1 binding loci to the chromatin by using whole genome ChIP-sequencing (ChIP-seq) technology (Fig. S6A). Rap1-null MEFs were used as control for RAP1 peak specificity. Over 21 and 17 million of uniquely mapped 36 base reads were collected for wild-type and Rap1-null cells, respectively (Table S14). In agreement with RAP1 binding to telomeres, we found a strong overrepresentation of raw 36bp sequences containing the telomeric TTAGGG5 repeat in WT Rap1 ChIP-seq compared to Rap1-null controls, which showed similar levels to input DNA (Fig. 7A). This difference was specific and not observed with non-telomeric repeats (Table S15; Fig. S6B). We also found perfect permutations of at least 2 telomeric repeats in the sequences of extra-telomeric RAP1 binding sites, in a frequency 32-fold higher (p-value < 2.2e-16) than that found in the mouse genome (Fig. 7B; Fig S6C). In total, between 2%-4% of all the genomic occurrences of telomeric repeats are occupied by RAP1 binding sites (Fig. S6D). When allowing 1 mismatch, the telomeric motif TTAGGG was found at least one time in up to 37.2% of the RAP1 peak sequences (Fig. S6E,F), suggesting that RAP1 binds to non-telomeric regions enriched in telomeric repeats.
In total, 30,398 non-telomeric binding sites (average peak width, 75bp) were identified upon comparing Rap1 WT and Rap1-null ChIP-seq datasets at 10% FDR level (Fig. S7A for peaks at 5% and 1% FDR levels). RAP1 sites were found throughout the genome (Fig. 7C), but showed a noticeable enrichment at the subtelomeric regions (17.45 sites/Mbp) (Fig. 7D) compared to the genome-wide density (11.71 sites/Mbp). The density of subtelomeric RAP1 peaks decreased proportionally to the distance from the telomere, suggesting a gradient of RAP1 binding from telomeres into the subtelomeric regions (Fig. 7E). The same trend was observed for the genomic distribution of (TTAGGG)2 repeats (Fig. S7B). Both signals showed a mild correlation (r=0.46; Fig. S7C), which was increased when considering regions closer to telomeres (r=0.89; Fig. S7C), suggesting a preference for repeat-mediated RAP1 binding at subtelomeres and the possibility of alternative binding mechanisms for the rest of the genome. The X chromosome showed the lowest density of RAP1-binding sites (2.65 RAP1 sites/Mbp; Fig. S7D), which was not attributable to differences in karyotype (all MEFs used were XY). Genes located in the X chromosome suffered less expression changes in Rap1-null cells compared to genes in other chromosomes (Fig. S7E), suggesting a correlation between RAP1 binding and gene expression changes.
Up to 22,384 individual RAP1 binding sites (73.63%) could be associated to genes, either mapping inside the gene structure (13,057 sites, 42.95%), or in a 10kbp window surrounding the gene transcription limits (9,327 sites, 30.68%) (Fig. 7F). The remaining 8,014 sites (26.36%) were considered intergenic (Fig. 7F). The increased density of RAP1 peaks around gene-coding regions resembles the genomic location patterns of known transcription factors. Given that coding genes and their 10kbp neighbor regions span about 47% of the mouse genome, this observation discards randomness in the topology of extra-telomeric RAP1 binding distribution (p-value < 2.2e-16), and supports that RAP1 may influence gene transcription. When exploring a 10kbp window surrounding gene transcription start sites (TSS) sites, up to 6,988 (22.98%) of the RAP1 sites could be associated to genes present in the gene expression array. Despite some strong RAP1 binding sites not associated to genes (Fig. 7H, peak in chromosome 2), 173 (31%) of the 558 deregulated (logFC>=1) genes presented 234 RAP1 binding sites mapping to their promoter regions (Table S16; Fig 7G). When comparing to the 4,980 non-deregulated genes with peaks, we found that the peak presence–gene deregulation dependency is significant (p-value = 0.0003), thus supporting a non-random relationship between RAP1 binding and transcriptional gene regulation. Significantly downregulated genes in Rap1-null cells containing RAP1 binding sites included Rap1 itself, Hic1 (downregulated 4-fold), Fgf5 (downregulated 5-fold), and Ctgf (downregulated 7-fold) (Fig. 7H; Table S16; Fig. 6A,B), suggesting that RAP1 acts as a transcriptional activator in these genes (see Fig. 8E). Together, these results suggest that RAP1 binding to extra-telomeric sites has an impact on transcriptional gene regulation.
To determine the sequence specificity of extra-telomeric mammalian RAP1 binding, we used the de-novo motif discovery algorithm Weeder 34 on a subset of the top over-represented RAP1-binding peaks (n=30). The highest-ranking motif was (TTAGGG)2 (Fig. 8A), probably due to presence of at least a perfect TTAGGG motif in up to 2.91% of the RAP1 peak sequences (n=884), and in up to 37.28% (n=11,332) when allowing 1 mismatch (Fig. S6F). A clear over-representation of the telomeric motif was found in the top 1000 and top 100 ranking peaks (Fig. 8B), with RAP1 peaks in chromosomes 11 and 17 showing more than 20 TTAGGG tandem repeats and the top ranking peak corresponding to a perfect (TTAGGG)8 repeat at chromosome 2 (Fig. 8B,C). This over-representation is evident in the peak sequences when compared to the aligned WT reads (Fig. S7F,G). When examining strongly deregulated genes, including Ctgf and Fgf5 genes (Table S17), Olfm1, Crabp1 and Angptl4, we found the telomeric consensus motif in their RAP1 binding peaks (Table S17; Fig. 8C for Olfm1, Crabp1 and Angptl4). We detected imperfect telomeric motifs in the peak sequences associated to half (51.44%) of DEGs (Table S17), a proportion similar to RAP1 peaks in non-deregulated genes (p=0.31). Given the significant association between RAP1 peaks and gene deregulation, these results suggests participation of RAP1 in alternative regulation mechanisms not mediated by telomeric repeat binding.
To explore non TTAGGG-rich putative RAP1 binding motifs we runned Weeder after discarting the top ranking peaks with ≥3 telomeric repeats. We obtained no clear consensus apart from heterogeneous A/T-rich sequences common for many transcription factors (Table S18).
To validate the ChiP-seq results, we chose a total of 14 RAP1-binding peaks located at both non-coding chromatin (Chrs. 2, 3, 17) and at strongly deregulated genes in wild-type versus Rap1-null MEFs (Fig. 8C,D; Table S15). As indicated by ChIP-qPCR, RAP1 showed binding to these regions and this was reduced in Rap1-null cells (Fig. 8D). No specific RAP1 binding was detected at chromatin regions that do not contain RAP1 binding sites (NC1 and NC2; Fig. 8D).
To address whether RAP1 binding to genes impacts their transcriptional activity, we cloned genomic fragments containing RAP1 binding sites at the promoter regions of Ctgf, Hic1 and Angptl4 upstream a minimal promoter driving luciferase expression (Fig. 8E). Upon transfection into wild-type and Rap1Δ/Δ cells, luciferase activity was significantly decreased in Rap1Δ/Δ cells compared to wild-type cells for all constructs tested while no differences were observed with the empty vector. These results suggest that these genomic regions have RAP1-dependent enhancer activity, supporting role for RAP1 in transcriptional regulation.
As RAP1 cannot bind chromatin directly and its recruited to telomeres by TRF2 3, we addressed whether TRF2 was able to bind to TTAGGG-rich extratelomeric RAP1 binding sites, such as those present in top ranking peaks at chromosomes 2 and 17. As determined by ChIP, TRF2 was bound to peaks in chromosomes 2 and 17 and this binding was independent of RAP1 (Fig. 8F). No significant TRF2 binding was detected at the RAP1 peak within the Crabp1 promoter (Fig. 8F), containing a degenerated telomeric repeat (Fig. 8C). Thus, RAP1 is likely to be recruited to TTAGGG-rich extratelomeric regions by TRF2 (ie, peaks at chromosomes 2 and 17), while other factors may meadiate RAP1 recruitment to degenerated telomeric repeats at the Crabp1 promoter.
Here, we describe cells and mice deleted for Rap1. While previous studies support a RAP1-dependent pathway for protection against telomere fusions35, 36, our current results suggest that chromatin-bound RAP1 does not significantly inhibit end-toend fusions in MEFs. Instead, Rap1-deficient MEFs show increased frequencies of multitelomeric signals associated to telomere fragility 14, 24, and elevated telomere recombination. This is in agreement with a recent report showing a role for RAP1 in preventing telomere recombination but not telomere fusions 13. In agreement with normal telomere capping, mice with targeted Rap1 deletion to stratified epithelia are viable and only show mild skin hyperpigmentation at adulhood. This is in marked contrast to 100% perinatal lethality and severe skin morphogenesis defects in similarly generated TRF1- and TPP1-null mice, with severe telomere capping defects 14, 15.
Rap1-deleted MEFs and mice showed a modest telomere shortening compared to controls, which was more apparent in Rap1-null proliferating tissues (ie, skin basal layer). This is in analogy to human diseases with presence of short telomeres due to telomerase mutations 37-41, also characterized by skin hyperpigmentation. As Rap1-null cells have normal telomerase expression, accelerated telomere shortening in Rap1-null epidermis may result from increased telomere recombination and fragility, although a role of RAP1 in TERT recuitment to telomeres cannot be discarted.
Yeast scRap1 and Trypanosoma brucei tpRap1 silence genes located near the telomeres, a phenomenon known as telomere position effect 6, 42. Here, we show preferential up-regulation of genes located in subtelomeric positions in Rap1-deficient cells compared to wild-type controls, thus suggesting a conserved role of RAP1 in subtelomeric gene silencing.
We show that Rap1-dependent transcriptional changes also include genes involved in cancer, cell adhesion, and metabolism. In particular, Rap1 deletion results in gene expression changes with a predicted negative impact on metabolism, in agreement an increased body weight phenotype in Rap1-null females. Of note, the transcriptional changes induced by Rap1 deletion are different from those associated to severe telomere dysfunction which involve activation of a stress response 43, thus although possible, it is unlikely that RAP1-dependent transcriptional effects at subtelomeres and elsewhere in the genome are due to mild telomere dysfuction induced by Rap1 deletion.
By using ChIP-seq technology, we demonstrate that RAP1 binds to telomeric repeats and to extra-telomeric sites preferentially throughout recognition of the (TTAGGG)2 consensus motif (Fig. 8E). Extra-telomeric RAP1 binding sites are enriched at subtelomeric regions and this enrichment decreases proportionally to the distance from the telomeres, providing a mechanism by which mammalian RAP1 can influence subtelomeric gene regulation. In addition, RAP1 binding sites are present in a significant percentage (31%) of the differentially regulated genes in Rap1-deficient cells, supporting a role of RAP1 in transcriptional gene regulation. RAP1 also binds to intergenic regions that contain many tandem telomeric repeats. The relevance of RAP1 binding to these internal telomeric repeats is unclear, although in analogy to telomeres, RAP1 may prevent recombination and fragility at these regions.
In summary, we describe both telomeric and non-telomeric roles for mammalian RAP1. Although non-telomeric roles have previously shown also for TERT and other shelterins 25, 44-48, our results provide first evidence of a mammalian shelterin component influencing gene transcription through binding to non-telomeric sites. We further show that TRF2 binds to RAP1 peaks containing many telomere repeats but not to those with degenerated repeats, in analogy to RAP1 recruitment to telomeres by TRF2 3. Future studies warrant identification of factors responsible of RAP1 recruitment to non TTAGGG-rich chromatin regions.
The strategy for disrupting the Rap1 locus was designed to conditionally delete E3 through a Cre-mediated excision. The targeting vector contains homology regions isogenic with the ES cell line used (129Sv/Pas). The short homology region (SA) harbors a 2.4 kb DNA fragment encompassing E2 and intron 2 (Fig. 1A). The long homology region (LA) is a 5.6 kb DNA fragment downstream the end of E3 (Fig. 1A). The central part contains E3 flanked by two loxP sites and a positive selection neomycin gene (PGK-Neo) flanked by two Frt sites (Fig. 1A). At the 3′-end of the LA a Diphteria Toxin (DTA) selection marker was cloned (Fig. 1A). The vector contains a unique NruI linearization site. The targeting vector was quality controlled by sequencing of the coding exons, the junctions between the homology arms and the selection cassettes. Sequencing showed no polymorphisms between the C57BL/6 and 129Sv/Pas genetic backgrounds within the isolated Rap1 sequences. The targeting vector was generated by genOway (www.genoway.com; Lyon, France).
129Sv/Pas ES cells were transfected with 40 μg of linearized targeting vector. Positive selection was performed by adding 200 μg/ML G418. Approximately 230 positive resistant clones were isolated and PCR screened for homologous recombination first at the 5′ end of the Rap1 locus. Six positive 5′ targeted ES cells were further investigated by PCR amplification over the 3′ long homology arm to amplify the region of the targeted locus containing the distal loxP site. Direct sequencing of the PCR products amplified revealed that 3 of out of 6 cloes contained the distal loxP site. They were verified by Southern blot analysis of AflII and PciI restricted genomic DNA using a 5′-internal and a 3′-external probes, respectively (data not shown). Chimeric mice were generated by microinjection of the 3 independently targeted ES clones into C57BL/6J host blastocyst. The resulting offspring showed a high level of chimerism as shown by coat color, and were mated to C57BL/6J mice to assess germ line transmission. The resulting heterozygous Rap1+/flox-Neo mice were then bred to transgenic mice expressing the Flpe recombinase 17 to induce excision of the Neo marker. The Rap1+/flox heterozygous mice were then intercrossed to generate Rap1flox/flox, and Rap1+/flox mice. Homozygous Rap1flox/flox mice were crossed with transgenic mice expressing the Cre recombinase under the control of the keratine 5 promoter 26 (Fig. 4A). Heterozygous Rap1+/Δ K5-Cre were crossed either to Rap1flox/flox or Rap1+/flox to generate Rap1Δ/ΔK5-Cre. The removal of exon 3 by Cre-mediated recombination was confirmed by PCR analysis using primers F and R (Fig. 1A,C,D and Fig. 4A). Amplification of the wild type, flox and knock-out alleles renders a 3.2 kb, 3.3 kb and 0.5 kb fragments, respectively.
The breeding to the F1 generation and characterization of heterozygous Rap1+/flox-Neo F1 animals was performed by geneOway (www.genoway.com; Lyon, France). All mice were generated and maintained at the Spanish National Cancer Centre (CNIO) under specific pathogen-free conditions in accordance with the recommendation of the Federation of European Laboratory Animal Science Associations.
Mice were fed either with standard chow diet (Harlan Teklad LM-485), or with a high fat diet (Research Diets 12451, 45% kJ from fat) starting at 7 weeks of age. Food intake was monitored by weighting the the consumed food every three days in individually caged animals.
Biological duplicates of ChIP samples (see above) were independently processed into sequencing libraries with a ChIP-Seq sample prep kit (Illumina) by following manufacturer instructions with some minor modification 49. Libraries were prepared from 85-135bp DNA fractions (excluding adaptor length). Input samples from both Rap1-null and wild type MEFs were pooled and sequenced as a single library. Libraries were sequenced in an Illumina Genome Analyzer IIx (GA2) single 36-base read run. Primary data was obtained by Pipeline 1.4 analysis (PL1.4, Illumina). Raw sequences were defined as reads passing purity filter before the genome alignment.
Genome alignment was performed with PL1.4 versus the latest mouse assembly (NCBIm37/mm9, April 2007) under default settings. These settings exploit the maximum mapping specificity allowed by the aligning algorithm. PL1.4 permits alignments with more than 2 errors for 36 base reads, but with no more than 2 errors in the first 32 bases. The best alignment among alternate candidate positions is eventually chosen based on quality scores. The experimental settings, sequences and analysis protocols of the ChIP-seq experiment have been deposited in GEO under the accession number GE20867.
Only the reads having a unique alignment in the reference genome where used for the peak detection which was performed using CisGenome v1.2 50. Briefly, uniquely aligned 36bp-length reads obtained in two independent GA2 runs were pooled into 3 datasets, corresponding to the Rap1 wild-type (WT) MEFs, Rap1-null MEFs (KO), and IP mockup. The software pipeline to analyze two-sample ChIP-seq experimental designs was applied WT reads as sample set and KO reads as negative control set using a 100bp-sliding window and a 10% FDR. To look for known genes in the neighborhood of the RAP1 binding sites, CisGenome searching method was executed against a database of mouse annotations, using a symmetrical window of 10kbp surrounding the TSS. The gene-annotated sites were cross-related with the gene labels in the microarray expression experiment by merging their expression profiles using BioConductor. Those merged genes with average logFC(WTvsKO)≥1 were considered as deregulated in the peak association experiments.
The dataset of ChIP-seq peaks was strongly filtered (minFDR ≤ 0.0001, maxFC≥6) to select the binding sites that were clearly represented. To derive any putative consensus motifs, the resulting 30 genomic sequences corresponding to the filtered peaks were processed with the de-novo motif discovery tool Weeder 1.3 34, as described in 51, 52. The motifs identified by the algorithm as highest ranking were selected, and the STAMP online tool 53 was used to represent the sequence logos 54 of the consensus sequences.
The pattern matching algorithms fuzznuc 55 and oligoMatch 56 were used to scan ChIP-seq binding sequences and mouse genomic sequence for permutations of telomeric repeats. To derive any putative non-TTAGGG consensus motifs, we discarded the peaks with 3 or more occurences of the telomeric repeats. We used STAMP 53 to determine the best match between the obtained motifs and known JASPAR and TRANSFAC v11.3 matrices.
The Chip-seq results were validated by q-PCR on pulled down DNA from an independently performed chip with a rabbit polyclonal anti-RAP1 (a gift from Dr. West, CRUK, UK). Oligos were designed to amplify the DNA fragment containing the peaks corresponding to Peak Rank 1 (chr2: 57,482,074-57,482,124), Peak Rank 4 (chr2: 28,040,118-28,040,149), Peak Rank 9 (chr3: 8,246,273-8,246,509), Peak Rank 14 (chr17: 53,291,729-53,291,957), Peak Rank 27 (chr17: 22,362,598-22,362,692), Peak Rank 41 (chr11: 3,091,929-3,092,050), Peak Rank 260 (chr17: 33,920,665-33,920,732), Peak Rank 561 (chr1: 40,496,711-40,496,796), Peak Rank 707 (chr13: 16,114,008-16,114,120), Peak Rank 764 (chr1: 12,701,710-12,701,939), Peak Rank 841 (chr19: 38,200,547-38,200,616), Peak Rank 2675 (chr10: 24,308,726-24,308,800), Peak Rank 8218 (chr11: 28,766,736-28,766,797), Peak Rank 29521 (chr9: 54,613,901-54,614,084). Genomic regions containing telomeric repeats that did not render any hit in Chip-seq analysis were choosen as negative controls; negative control 1 (chr1:121,265,202-121,265,386) and negative control 2 (chr15:54,909,216-54,909,375). The primer sequences are available upon request. The amplification levels of the above-mentioned fragments were analyzed in the input DNA in each case for normalization and in the pulled down DNA. The relative level of each fragment was determined by calculating ΔCt values between the levels obtained in input DNA and that of the pulled down DNA. The results were normalized to wild type levels.
TRF2 binding to Peak Rank 1, Peak Rank 14 and to Peak Rank 29521 corresponding to Crabp1 promoter was performed as described above using pulled down DNA with polyclonal TRF2 antibody (a gift from Dr. West, CRUK, UK). The results were normalized to input levels.
DNA fragments (~300 bp) harbouring Peak Rank 2675 (chr10: 24,308,726-24,308,800), Peak Rank 19335 (chr11: 74,972,250-74,972,307) and Peak Rank 260 (chr17: 33,920,665-33,920,732) located at CTGF, HIC1 and ANGPTL4 promoter region, respectively, were PCR amplified from mouse genomic DNA. Forward and reverse primers contained a XhoI and a HindIII and their sequence are available upon request. The fragments were cloned into XhoI-HindIII sites of pGL4.28 vector containing a minimal promoter driving luciferase (Promega) and sequence verified. LT-sv40 immortalized Rap1+/+ and Rap1Δ/Δ MEF were transfected with the reporter constructs by using Fugene (Roche). A plasmid (pGL4.75, promega) containing a CMV promoter driving Renilla luciferase was cotransfected as an internal control. Cells were harvested 48 hours after transfection, and the luciferase activities of cell lysates were measured by using Dual-luciferase Reporter Assay System (Promega).
A t-student test was used to calculate the statistical significance of the observed differences in the percentage of RAP1 positive cells, γH2AX foci, 53BP1 foci and chromosomal aberrations. Wilcoxon-Mann-Whitney rank sum test was used to calculate statistical significance of the observed differences in the mean telomere length. A two-tailed Wilcoxon-rank test was used to calculate the statistical significance of the differences in gene expression levels amongst subtelomeric and non-subtelomeric genes. To determine the gene expression levels trend in subtelomeric regions between classes under comparison we applied a one-tailed Wilcoxon-rank test. P-values<0.05 were considered statistically significant. Fisher Exact test was used to estimate the differences in the presence of repetitive sequences in the raw reads. To test whether is possible to find a binging peak located in or near the genic regions just by random chance we performed an exact binomial test. Pearson’s Chi-square test was used to check the dependency between the presence of RAP1 binding peaks and expression deregulation of their associated genes. Freely distributed R software was employed for these calculations (http://www.r-project.org/).
We are indebted to L. Harrington for Tert-deficient MEFs and S. West for RAP1 antibodies, and R. Serrano for animal care. P.M. is funded by a “Ramón y Cajal” from the Spanish Ministry of Innovation and Science. M. A. Blasco’s laboratory is funded by the Spanish Ministry of Innovation and Science, the European Union (FP7-Genica, Telomarker), the European Reseach Council (ERC Advance Grants), the Spanish Association Against Cancer (AECC), and the Körber European Science Award to M.A. Blasco. The work in M. Tarsounas Labroatory is funded by Cancer Research UK.
Competing Interests The authors declare that have no competing finantial interests.
Accession Numbers The microarray and ChIP-seq datasets have been deposited in the GEO database under the accession numbers GSE19011 and GSE20867, respectively.
Additional Methods are described in the Supplementary Methods section in Supplementary Information