|Home | About | Journals | Submit | Contact Us | Français|
Author Information Sequencing files deposited at GEO (accession codes GSE22411 and GSE21673). Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Readers are welcome to comment on the online version of this article at www.nature.com/nature.
Multiple pathways prevent DNA replication from occurring more than once per cell cycle1. These pathways block re-replication by strictly controlling the activity of pre-replication complexes, which assemble at specific sites in the genome called origins. Here we show that mutations in the homologous histone 3 lysine 27 (H3K27) monomethyltransferases, ARABIDOPSIS TRITHORAX-RELATED PROTEIN5 (ATXR5) and ATXR6, lead to re-replication of specific genomic locations. The vast majority of these locations correspond to transposons and other repetitive and silent elements of the Arabidopsis genome. These sites also correspond to high levels of H3K27 monomethylation, and mutation of the catalytic SET domain is sufficient to cause the re-replication defect. Mutation of ATXR5 and ATXR6 also causes upregulation of transposon expression and has pleiotropic effects on plant development. These results uncover a novel pathway that prevents over-replication of heterochromatin in Arabidopsis.
We previously characterized two redundant histone methyltransferase genes, ATXR5 and ATXR6, and demonstrated that the ATXR5 and ATXR6 proteins show H3K27 monomethylation (H3K27m1) activity in vitro, and that an atxr5 atxr6 double mutant shows a reduction of H3K27me1 in vivo2. The atxr5 atxr6 double mutant exhibits pleiotropic defects in plant development, including smaller misshapen leaves2. Overexpression of ATXR5 or ATXR6 also causes morphological defects and male sterility3. Furthermore, the double mutant displays reactivation of the expression of a variety of both DNA transposons and retrotransposons2. Interestingly, atxr5 atxr6 did not disturb DNA methylation or histone H3K9 dimethylation (H3K9me2, a key repressive histone modification correlated with DNA methylation4, 5, 6), suggesting that ATXR5 and ATXR6 act via a novel pathway to maintain gene silencing. Previous work also provided hints that ATXR5 and ATXR6 show links with DNA replication. ATXR5 and ATXR6 expression is regulated by the cell cycle, with expression peaking just before DNA replication3 and ATXR6 expression is strongly coregulated with CDT1, ORC2 and other DNA replication proteins7. In addition, ATXR5 and ATXR6 contain PCNA-interacting-protein (PIP) motifs and have been shown to interact with the two PROLIFERATING CELL NUCLEAR ANTIGEN (PCNA) proteins in Arabidopsis (AtPCNA1 and AtPCNA2)3. PCNA interacts with DNA polymerase and serves as a general loading platform for many proteins involved in diverse processes occurring at chromatin8. In this report, we show that ATXR5 and ATXR6 are critical factors that act in a novel pathway to suppress DNA re-replication, especially in heterochromatic regions of the Arabidopsis genome.
To determine if ATXR5 and/or ATXR6 have a role in DNA replication, we analyzed the DNA content of leaf nuclei by flow cytometry. Leaves are well suited for assessing DNA replication defects as they undergo both mitosis and endoreduplication (genome duplication without mitosis), which is responsible for the widespread polyploidy observed in mature leaf tissue9. Nuclei were extracted from mature rosette leaves of wild-type Columbia (Col), atxr5, atxr6, and atxr5 atxr6 plants, stained with propidium iodide (PI), and analyzed for DNA content. Col, atxr5, and atxr6 all showed well-resolved populations of 2C, 4C, 8C, 16C, and 32C nuclei (Fig. 1a). Thus atxr5 and atxr6 single mutations do not have an impact on DNA replication in leaves. In contrast, the 8C and 16C peaks for atxr5 atxr6 were much broader and skewed to the right, suggesting that many 8C and 16C nuclei have higher DNA contents than the corresponding wild-type nuclei (Fig. 1a). The distribution of nuclei between endoreduplication levels was similar between wild type and atxr5 atxr6 double mutants (Fig. 1b), suggesting that the primary defect in atxr5 atxr6 plants is in the fidelity of S-phase progression rather than the number of rounds of endoreduplication. This is in contrast to mutations previously reported to affect the level of endoreduplication, but not S-phase fidelity10.
Our previous work has shown that ~65% of atxr5 atxr6 nuclei show significant decondensation of constitutive heterochromatin (i.e., chromocenters) (Fig. 1c)2. Consistent with the flow cytometry results showing that the DNA content phenotype of atxr5 atxr6 is observed most strongly in 8C and 16C nuclei, we found by microscopic analysis of sorted nuclei that the heterochromatic decondensation defect was also more extreme in these higher ploidy nuclei (Fig. 1c,d). As a control, we also analyzed the DNA content of decrease in dna methylation1 (ddm1) plants which also show very strong chromocenter decondensation defects, as well as reduced DNA methylation and massively reactivated transposons11-13. The flow cytometry profile from extracted nuclei from ddm1-2 leaves was similar to that of wild-type plants (Fig. 1a). These results show that the aberrant flow cytometry profiles observed in atxr5 atxr6 are not simply a result of chromatin decondensation defects and/or transposon derepression.
To test the hypothesis that there is indeed extra DNA in atxr5 atxr6 mutants, we used an Illumina Genome Analyzer II to sequence genomic DNA from sorted nuclei (2C, 4C, 8C, and 16C) of both wild-type and atxr5 atxr6 plants. A total of 84.9 million uniquely mapping 36-nucleotide reads were mapped to the Arabidopsis thaliana genome allowing up to two mismatches. We examined the distribution of genomic DNA across all chromosomes by plotting the density of reads in non-overlapping 100kb bins. For wild type the ratio of 4C, 8C, and 16C to 2C sequence reads were uniform across the genome, showing the genome is uniformly endoreduplicated in wild-type Arabidopsis (Fig. 1e). Strikingly however, in atxr5 atxr6 mutants, we observed an enrichment of reads in the pericentromeric heterochromatin in 4C, 8C and 16C compared to 2C, indicating that heterochromatin is over-replicated in 4C, 8C and 16C nuclei compared to 2C nuclei (Fig. 1f), and that the over-replication is more severe in nuclei with higher ploidy levels.
A comparison of the distribution of reads from atxr5 atxr6 mutants with wild type showed that even in the 2C nuclei, atxr5 atxr6 mutants show over-replication of pericentromeric heterochromatin, though to a lower extent than in nuclei of higher ploidy levels (Supplementary Fig. 1a). Relative to wild type, atxr5 atxr6 mutants showed a 2.9%, 11.0%, 29.1%, 28.4% increase in reads mapping to pericentromeric heterochromatin in 2C, 4C, 8C, and 16C nuclei respectively (Supplementary Fig. 1b). Sites of over-replication were well correlated at the different ploidy levels. For instance the Pearson correlation between 8C and 16C atxr5 atxr6 nuclei was 0.84, indicating that the same sites are over-replicating (Supplementary Fig. 1c). These results show that atxr5 atxr6 mutants exhibit over-replication of pericentromeric heterochromatin in both mitotic and endocycling cells, with progressively stronger defects observed in nuclei with higher ploidy levels.
To examine over-replication in atxr5 atxr6 mutants at higher resolution, sequence reads were grouped and analyzed in 200 base pair non-overlapping bins. We found that over-replication of pericentromeric heterochromatin is the result of the over-replication of many densely spaced, but distinct loci (Fig. 2a). We also observed localized over-replication in small regions of the euchromatic arms of chromosomes (Fig. 2b). These small regions of over-replication were highly enriched in transposons and other repeat elements. Using the BLOC algorithm (see Supplementary Methods)14, we identified 407 sites of over-replication in the arms of atxr5 atxr6 chromosomes (Supplementary Table 1). The over-replicating regions were relatively small; 94% of regions were smaller than 25kb, with a median size of 10.4kb (Supplementary Fig. 2a). The majority (80%) overlapped with previously defined H3K9me2 regions, a mark with is strongly correlated with DNA methylation and gene silencing6 (Supplementary Fig. 2b). Thus, the regions that over-replicate in atxr5 atxr6 mutants primarily consist of transposons and silent elements of the Arabidopsis genome. Over-replication was confirmed by performing qPCR on defined sites (Supplementary Fig. 3). Elements that are transcriptionally reactivated in atxr5 atxr6 mutants (TSI, Ta3, CACTA)2 were found to be over-replicated, suggesting a positive correlation between transposon reactivation and over-replication.
Re-replication is a well-known mechanism by which DNA is known to over-replicate and results when DNA replication is initiated from an origin multiple times during a single S phase1. Presumably because recently replicated chromatin is less compact, secondary replication forks move faster than primary forks, and collisions of the multiple forks result in successively smaller fragments of DNA reiteratively produced from the origin15 (Fig. 2c). This model predicts that sequences in the center of the origin will be the most highly over-replicated and that over-replication should drop off symmetrically on either side of the origin. To determine if over-replication in the atxr 5atxr6 mutant is consistent with re-replication, plots of sequencing reads averaged over the over-replicated regions were generated (Fig. 2d). In contrast to wild type, in which sequencing reads were uniformly distributed, atxr5 atxr6 mutants showed a bilaterally symmetrical distribution of reads, with the highest density of reads in the center of the over-replicated regions (Fig. 2d). These results suggest that the extra DNA in atxr5 atxr6 mutants is a result of repeated replication from defined sites.
We next examined whether chromatin or naked DNA is being re-replicated in atxr5 atxr6 mutants. To test this, we performed chromatin-immunoprecipitation of unmodified histone H3 followed by Illumina sequencing (ChIP-seq) on wild-type and atxr5 atxr6 mutants. Compared to wild type, H3 ChIP-seq reads in atxr5 atxr6 mutants were enriched in the pericentromeric heterochromatin to a similar extent as was the input genomic DNA (Fig. 2e and Supplementary Fig. 2c). This result suggests that chromatin (DNA and associated histones) is re-replicated in atxr5 atxr6 mutants. Our data also suggest that the re-replicated DNA is properly methylated. If the re-replicated DNA was unmethylated, the % methylation in atxr5 atxr6 mutants would be predicted to be lower than in wild type. However, we have previously shown that the % DNA methylation in atxr5 atxr6 mutant leaves (where a significant amount of re-replication was observed) was the same as in wild-type2, which suggests that the re-replicated DNA is properly methylated. In addition, to determine whether re-replicated DNA was stably associated with the chromosomes, we performed qPCR on size fractionated DNA (Supplementary Fig. 4). We found that the extra DNA could be detected in the high molecular weight DNA fraction, suggesting that at least part of the re-replicated DNA is stably associated with the chromosome. Being associated with chromosomes, rather than being extrachromosomal fragments, may help explain the stability of re-replicating DNA fragments present in 3-4 week leaf cells.
Because ATXR5 and ATXR6 catalyze H3K27me12, we wanted to examine whether the spatial distribution of H3K27me1 overlaps with re-replicating regions. Immunolocalization suggests that H3K27me1 is a heterochromatic mark enriched in chromocenters2, 16-18. A detailed global map of H3K27me1, however, has not been reported. We therefore profiled H3K27me1 genome-wide using ChIP-seq. Consistent with the re-replication of pericentromeric heterochromatin in atxr5 atxr6 (Fig. 1f), we found that H3K27me1 was strongly enriched in pericentromeric heterochromatin (Fig. 3a). We also observed H3K27me1 in the coding regions of protein-coding genes and found that the amount of H3K27me1 was anticorrelated with gene expression levels (Fig. 3b, and Supplementary Fig. 5a). Together, these results support a role for H3K27me1 in gene silencing.
To gain additional evidence for a correlation between H3K27me1 and re-replication in atxr5 atxr6 mutants, we examined the dispersed re-replicating regions in the arms of the chromosomes. We found that H3K27me1 ChIP-seq reads were significantly enriched in these regions compared to randomly selected control regions (permutation test, P<10-6) (Fig. 3c). In addition, plots of the ratio of H3K27me1 to H3 ChIP-seq reads averaged over these re-replicating regions showed strong enrichment of H3K27me1, confirming a positive correlation of H3K27me1 with sites that re-replicate in atxr5 atxr6 mutants (Fig. 3d and Supplementary Fig. 5b).
Given that H3K27me1 levels correlate with the re-replicated regions of atxr5 atxr6 mutants, an interesting question concerns the mechanism by which the spatial distribution of H3K27me1 is established. ATXR5 and ATXR6 both contain PHD domains, which have been shown in multiple species to mediate interactions with methylated or unmethylated forms of histone H319. We performed in vitro binding assays with various H3 peptides using GST-tagged PHD domains of ATXR5 and ATXR6. The PHD domains of ATXR5 and ATXR6 bound strongly to an unmethylated peptide corresponding to amino acids 1-21 of H3 (Fig. 3e). This binding was unaffected by mono-, di-, or trimethylation at H3K9, however, binding was strongly reduced by increasing levels of H3K4 methylation. Thus the PHD domains of ATXR5 and ATXR6 bound most strongly to H3 unmethylated at K4 (H3K4me0). Consistent with the hypothesis that binding of the ATXR5 and ATXR6 PHD domains to H3K4me0 chromatin is helping to guide H3K27 monomethylation activity, we observed a strong anticorrelation between H3K4 methylation20 and H3K27me1 within genes and in the genome at large (Fig. 3f, Supplementary Fig. 5c).
Because loss of ATXR5 and ATXR6 leads to lower levels of H3K27me12, it is possible that depletion of this mark is causing re-replication in atxr5 atxr6 mutants. One prediction from this model is that the PHD- and SET domains of ATXR5 and ATXR6 would be essential to prevent re-replication, as they are responsible for binding and methylating H3, respectively (Fig. 4a). To investigate this, we first created a genomic construct that expresses ATXR6 under its own promoter and confirmed that it can rescue (>95% of T1 transformed plants analyzed) the re-replication phenotype of atxr5 atxr6 plants (Fig. 4b). We then made PIP-, PHD-, and SET-mutant ATXR6 constructs by inserting point mutations designed to disrupt the activity of each functional element (Fig. 4a). Yeast-two-hybrid analysis and in vitro histone-peptide-binding and methyltransferase assays were used to confirm disruption of the PIP motif, PHD-, and SET-domain activities, respectively (Supplementary Fig. 6). Analysis of T1 plants transformed with each of the mutated ATXR6 constructs showed that the re-replication phenotype was never rescued by constructs containing the mutated PIP motif, PHD- or SET domains (Fig. 4b) (n>20). These results show that the PIP motif, PHD and SET domains are all required for ATXR6 activity and suggest that depletion of H3K27me1 in the atxr5 atxr6 double mutant is likely responsible for the re-replication phenotype. Consistent with this interpretation, we found that the restoration of H3K27me1 levels also required the wild-type PIP motif, PHD and SET domains (Supplementary Fig. 7). Furthermore, only the wild-type construct rescued chromatin decondensation and loss of gene silencing at defect seen in atxr5 atxr6 mutants (Fig. 4c,d). These results indicate that the three functional elements of ATXR6 contribute to the prevention of re-replication, chromatin decondensation, and loss of gene silencing.
In summary, our results suggest that ATXR5 and ATXR6 are components of a novel pathway required to suppress re-replication in Arabidopsis. Remarkably most of the re-replicating sites in atxr5 atxr6 mutants correspond to silent heterochromatin, which is composed mostly of transposon sequences. It is tempting to speculate that the ATXR5 ATXR6 system may have evolved to suppress excess DNA replication of transposon sequences that would otherwise result in transposon reactivation. Conversely, transposons are remarkable in requiring both the typical repressive modifications such as H3K9me2 and DNA methylation, as well as the novel ATXR5/6 H3K27me1 pathway for transcriptional suppression.
FACS was used to generate flow cytometry profiles of leaf nuclei from Col and atxr5 atxr6 and to sort nuclei based on endoreduplication level (2C, 4C, 8C, and 16C). Genomic DNA isolated from sorted nuclei was sequenced using an Illumina Genome Analyzer II. SeqMap (11) was used to map sequencing reads to the Arabidopsis genome. ChIP-seq using H3 and H3K27me1 antibodies were sequenced and analyzed in a similar fashion. In vitro binding assays utilized biotinylated H3 peptides (Millipore, Billerica, MA).
We thank G. Lambert and D. Galbraith (University of Arizona) for assistance with flow cytometry, Y. Bernatavichute for assistance with ChIP experiments, and M.Pellegrini and S. Cokus for advice on data analyses. Y.J. was supported by a fellowship from Le Fonds Québécois de la Recherche sur la Nature et les Technologies (FQRNT). S.F. is a Howard Hughes Medical Institute Fellow of the Life Science Research Foundation. Research in the Michaels’ lab supported by grants from the National Institutes of Health (GM075060), the Indiana METACyt Initiative of Indiana University, and the Lilly Endowment, Inc. S.E.J. is an investigator of the Howard Hughes Medical Institute.
atxr5 (SALK_130607) and atxr6 (SAIL_240_H01) in the Columbia genetic background were obtained from the Arabidopsis Biological Resource Center. The ddm1-2 seeds have described previously (1, 2). Plants were grown under cool-white fluorescent light (~100 μmol m-2 s-1) under long-day conditions (16 h of light followed by 8 h of darkness).
Plant tissue was chopped with a razor blade in 500 μl of Galbraith buffer (45 mM MgCl2, 20mM MOPS, 30 mM sodium citrate, 0.1% Triton X-100) containing 20 μg of RNase A. The lysate was filtered through a 40 μm cell strainer (BD Falcon) and propidium iodide was added to a final concentration of 20 μg/ml. For mature leaves, nuclei were extracted from rosette leaves 3 and 4 from 4-week-old plants; for young leaves, nuclei were extracted from rosette leaves 1 and 2 from 9-day-old plants. Flow cytometry profiles were obtained on a BD FACSCalibur (Becton Dickinson, San Jose, CA) based upon PI and 90° side angle scatter on a logarithmic scale due to the small size of the nuclei. Analysis was performed with CellQuest Pro software (Becton Dickinson, San Jose, CA). For nuclei sorting, 2.5 g of mature rosette leaves were collected from 4-week-old plants, chopped in 5 ml of Galbraith buffer containing 200 μg of RNase A, filtered and stained with propidium iodide. The procedure for sorting was based on parameters similar to the FACSCalibur on a BD FACS Aria II, using a 85μm nozzle with sheath pressure at 35 psi. Analysis was performed with FACSDiva version 6.1.1 (Becton Dickinson, San Jose, CA). For sequencing, 400,000 nuclei of each ploidy (2C, 4C, 8C and 16C) were isolated from Col and atxr5 atxr6 leaves at a threshold rate of approximately 500 events/sec, a flow rate of 2.5 and a variable sort rate dependent upon the peak sorted. Genomic DNA was extracted from the sorted nuclei using the PicoPure DNA Extraction Kit (Molecular Devices, Sunnyvale, CA) according to the manufacturer's recommendations.
For analysis of sorted nuclei by microscopy, 0.25 g of leaves were fixed for 20 min in 4% formaldehyde in Tris buffer (10 mM Tris-HCl pH 7.5, 10 mM EDTA, 100 mM NaCl) and washed with Tris buffer 2 × 10 min. The leaves were then chopped in Galbraith buffer, filtered and stained with 4 μg/ml DAPI. Two thousand nuclei of each ploidy were sorted directly onto microscope slides. Coverslips were then mounted using Vectashield mounting medium with DAPI (Vector Laboratories, Burlingame, CA) and sealed with clear nail polish. 30 nuclei for each ploidy type were analyzed for three biological samples. Flow cytometry and nuclei sorting were performed at the Indiana University Flow Cytometry Core Facility
The procedures were performed as described previously (3). Microscopy was performed at the Indiana University-Bloomington Light Microscopy Imaging Center.
To make the ATXR6 constructs used to transform A. thaliana, the promoter (293 bp upstream of start codon) and gene (exons and introns) were amplified by PCR and cloned into pENTR/D (Invitrogen, Carlsbad, CA). Point mutations in the PHD domain, SET domain, and the PIP motif were made using pENTR/D-ATXR6 as template by site-directed mutagenesis. For the PHD-domain mutant, leucine 49 was replaced with tryptophan (L49W). This mutation is based on the structure of the PHD domain of the mammalian protein BHC80, which was shown to rely on an equivalent residue (M502) for preferentially binding H3K4me0 (4). To disrupt the function of the SET domain, we changed tyrosine 243 to asparagine (Y243N). This tyrosine is part of the conserved YXG motif of SET-domain proteins, which is responsible for orienting the ε-amino group for methyl transfer to occur on lysine (5). An equivalent mutation (Y655N) in Drosphila ENHANCER OF ZESTE (E(z)) was shown to abolish H3K27 methylation without affecting folding of the protein (6). Finally, glutamine 92, isoleucine 95, phenylalanine 98, and phenylalanine 99 of the conserved PIP motif (QTKIIDFF) of ATXR6 were replaced with alanine. The resulting ATXR6 constructs were subcloned first into pEG302 (7) using the Gateway technology to acquire a C-terminal FLAG-epitope sequence, then into pMDC30 (8) by restriction digest of the pEG302 vectors with SpeI and SbfI.
pGEX-6P was used to clone the PHD domains (a.a. 25–103, ATXR6; a.a. 57–133, ATXR5) and PHD-SET domains (aa residues 25–349, ATXR6) for the in vitro binding and methyltransferase assays, respectively. The point mutations in the PHD (L49W) and SET (Y243N) domains were introduced in the pGEX vectors by site-directed mutagenesis.
For the yeast two-hybrid assay, the coding sequences of ATXR6, ATXR6(pip), and AtPCNA1 (At1g07370) were cloned into pENTR/D, then subcloned into pDEST32 or pDEST22 using the Gateway system.
Biotinylated peptides were obtained from Millipore, Billerica, MA. The procedure for this assay has been described previously (9). The binding buffer contained 250 mM NaCl. Histone methyltransferase assays were preformed as previously described (3).
pDEST22-AtPCNA1 was co-transformed with each of the two pDEST32-ATXR6 vectors in Saccharamocyes cerevisiae (strain MaV203, Invitrogen, Carlsbad, CA). As controls, pDEST22-AtPCNA1 and pDEST32-ATXR6/ATXR6(pip) were also used for co-transformation with empty pDEST32- and pDEST22 vectors, respectively. Five independent colonies from each transformation were selected and streaked on SC –Leu–Trp plates. The colonies were then replica plated on SC –Leu–Trp plates containing various concentrations (25, 50 or 100 mM) of 3-Amino-1,2,4-Triazole (3AT), and grown for 3-4 days at 30°C.
ChIP was performed on crosslinked 3-week-old plants as previously described (10) using α-H3 (Abcam #1791) and α-H3K27me1 (Upstate #07-448).
50,000 16C nuclei were prepared from rosette leaves of 28-day-old plants (Col and atxr5 atxr6), as described above. For sorting, same 16N gate size was used for both samples. Genomic DNA was extracted from the sorted nuclei using the PicoPure DNA Extraction Kit (Molecular Devices, Sunnyvale, CA) according to the manufacturer's recommendations. qPCR was performed using Brilliant II SYBR green QPCR master mix(Agilent), also according to the manufacturer's recommendations.
Four 3-week-old leaves were ground and extracted in 400 μL extraction buffer (sorbitol 350 mM, Tris-HCl pH7.5 100mM, EDTA 5 mM) followed by lysis in 400 μL of lysis buffer (Tris-HCl pH7.5 200mM, EDTA 50mM, NaCl 2M, CTAB 2%) plus 27 μL of 10% sarkosyl. After a 30 minute incubation at 65°C, DNA was extracted with chloroform/isoamyl alcohol, precipitated and resuspended in 100 μL of extraction buffer including RNase A (Qiagen). Around 10 μg of this DNA was run in a 1% low melting point agarose gel (Lonza) for both Col0 and atxr5 atxr6 and the band for the genomic DNA was cut out of the gel (BIG) along with the rest of the lane below that band (SMALL, DNA<12 Kb).
For the recovery of the BIG fraction of DNA from the gel we diluted it 5 times with TE and incubated at 65°C until the agarose melted. We then performed serial extractions with phenol, phenol/chloroform and chloroform followed by isopropanol precipitation and resuspension in 50 μL of TE.
For the recovery of the SMALL fraction of DNA from the gel we used Qiagen gel extraction kit and followed the manufacturer's instructions. DNA was eluted in 50 μL of TE. qPCR was performed using the IQ -SYBR Green Supermix from Bio-Rad, a Stratagene MX3005P qPCR system.
Illumina libraries for genomic DNA extracted from sorted nuclei and ChIP samples were made following manufacturer instructions. The libraries were sequenced using Illumina Genome Analyzer II following manufacturer instructions, producing reads of 36bp in length.
Sequenced reads were based-called using the standard Illumina software. We used SeqMap (11) to align the reads to the Arabidopsis thaliana genome, allowing up to 2 mismatches. Identical reads were collapsed into single reads. We used two different mapping strategies for all of our data sets: (1) keeping reads that uniquely map to the genome and (2) remapping the reads that map to multiple locations in the genome, giving each read a weight. In (2), if a read mapped to x places in the genome, the read was given a score of 1/x, hence uniquely mapping reads got a score of 1. (2) was necessary since the majority of re-replicating regions were present in heterochromatic regions, and were used to define regions of re-replication. Since many heterochromatic regions are highly repetitive, we found that uniquely mapping the reads results in no signal in those regions.
Since 36 base pair reads represented the ends of the DNA fragments in the library, for the analysis, we extended the read so that the data represents the actual DNA fragments of the libraries. The lengths of extensions were determined based on the distribution of the sizes of DNA fragments in the library. Each base pair of the read was given a score of 1 (or 1/x). Therefore if a certain nucleotide in the genome had x uniquely mapping fragments overlapping, that nucleotide got a score of x. Data in (1) were normalized to total number of uniquely mapping reads in wild-type, whereas in (2), data were normalized to the number of uniquely mapping reads + the number of reads that map to multiple locations.
Re-replicating regions were defined by using BLOC (12). Scores for atxr5 atxr6 log2(16C/2C) was calculated in 60bp bins, Z-score transformed, and then an average Z-score cutoff of 0.3 was applied for BLOC.
Supplementary Information is linked to the online version of the paper at www.nature.com/nature.