Epigenetic Instability due to Defective Replication of Structured DNA

Summary The accurate propagation of histone marks during chromosomal replication is proposed to rely on the tight coupling of replication with the recycling of parental histones to the daughter strands. Here, we show in the avian cell line DT40 that REV1, a key regulator of DNA translesion synthesis at the replication fork, is required for the maintenance of repressive chromatin marks and gene silencing in the vicinity of DNA capable of forming G-quadruplex (G4) structures. We demonstrate a previously unappreciated requirement for REV1 in replication of G4 forming sequences and show that transplanting a G4 forming sequence into a silent locus leads to its derepression in REV1-deficient cells. Together, our observations support a model in which failure to maintain processive DNA replication at G4 DNA in REV1-deficient cells leads to uncoupling of DNA synthesis from histone recycling, resulting in localized loss of repressive chromatin through biased incorporation of newly synthesized histones.


Chromatin immunoprecipitation
Modifications to the ChIP procedure of Aparicio et al. (Aparicio et al., 2005). Lysates were prepared and the DNA was fragmented by sonicating to a size of <1.5 kb with an average of 400-600 bp as shown by analysis on a 1.5 % agarose gel. An amount of crosslinked lysate calculated to contain approximately 25-50 µg of DNA was diluted to a total volume of 500 µl with FA lysis buffer (Aparicio et al., 2005) before incubation overnight at 4 ˚C with the ChIP antibodies (see below). The antibody complexes were then precipitated by incubation with 30 µl Protein G magnetic beads (New England Biolabs) for 2 hours at 4˚C. After sequential washing of the beads with FA lysis buffer, FA lysis buffer with 0.5M NaCl, a LiCl-based chromatin immunoprecipitation wash buffer (Aparicio et al., 2005) and TE buffer, complexes were eluted by shaking in an Eppendorf Thermomixer at 65 ˚C for 30 minutes in 100 µl ChIP elution buffer (1 % SDS, 100 mM NaCO 3 ). Following this, the mixture was treated with 5 µg Proteinase K (1 mg/ml solution, Sigma), for 2 hours at 42 ˚C and cross-links were reversed by incubating for 4 hours at 65 ˚C. Finally, DNA was precipitated using sodium acetate and ethanol and analysed by quantitative PCR using primers listed in Table S2.

Computer simulation of histone recycling
The simulation uses as its biological basis a similar model to that used by Thon and colleagues (Dodd et al., 2007), but employs stochastic simulation rather than one based on differential equations. For simplicity, only three states of nucleosomes are considered, modified, unmodified and naïve, the latter being the state of a newly synthesised histone. The simulation uses a uniform tract of stably modified histones as its template and histone replication is semiconservative. Thus, for every two nucleosomes deposited on the newly synthesised DNA, one is from the parental duplex and one is newly synthesised. However, the order in which they are arranged is random, reflecting the stochastic nature of histone segregation onto the daughter strands. Once this is complete, the modified histones' marks are copied onto adjacent naïve nucleosomes. Additionally there is a finite probability that the modified mark can spread more than one nucleosome away (two-place copying), thus endowing the system with robustness in the face of errors. The model can 'stall' replication with a defined per nucleosome probability and, to test our hypothesis for histone deposition during the restoration of post-replicative gaps, it deposits a tract of naïve histones in the gap, the length of the gap being another variable parameter. The implementation is divided into three stages: Stage 1-Replication Nucleosomes are removed from the template. For every two successive nucleosomes removed, one retains its original mark, and the other assumes an unmodified 'naïve' state. This process is subject to random error limited to one nucleosome, given by the parameter 'probability of mark misincorporation'. If a stall occurs, however, with probability determined by the parameters 'probability of roaming stall' or 'probability of fixed stall', the simulation inserts an integral number of naïve histones with the parameters 'gap length' and 'gap length variance' specifying Gaussian mean and variance respectively.
Stage 2 -Mark copying. The final state of each naïve nucleosome is determined by the states of the adjacent marked nucleosomes. 'Stop' and 'go' marks, corresponding to repressed and active nucleosome states respectively, can each copy onto the naïve nucleosomes with probability determined by the parameters 'probability of stop copying' and 'probability of go copying'. Additionally, the parameter rho determines the probability of a mark copying to a naïve nucleosome two places away: rho is the ratio of probability of one place copying to probability of two place copying.
Stage 3 -Completion The nucleosomes are returned to the template. At this point, any nucleosome that has not obtained a modification, i.e. remains naïve, will acquire the modification 'go', to reflect persistence of H4 N terminal acetylation. The simulation is now ready for another round of division.  Figure 1A and results are normalised to total H3, but not then normalised again to the hypersensitive site as no information exists on the global pattern of H3K56ac in vertebrates. Error bars represent standard error of the mean. Figure S2. In Situ Disruption of the BRCT Domain of REV1 in DT40, Related to Figure 4. We have previously reported that reconstitution of rev1 DT40 cells with a construct expressing human REV1 lacking the N terminal 332 amino acids was able to fully complement phenotypes of the mutant including damage sensitivity and defective replication fork progression after DNA damage, suggesting that the BRCT domain is dispensable for DNA damage tolerance (Edmunds et al., 2008;Ross et al., 2005). However, we suggested that the discrepancy between these observations and the apparently essential role of the BRCT domain in yeast (Nelson et al., 2000) might be explained by our ectopic expression bypassing the requirement for the BRCT domain (Ross et al., 2005). To address this we created a DT40 rev1 BRCT mutant in which the protein is expressed under its natural promoter.
A. Alignment of the chicken (Gg) REV1 BRCT domain with the BRCT domains of yeast (Sc) REV1 and human (Hs) BRCA1, XRCC1 and 53BP1 made with CDD (Marchler-Bauer et al., 2009). The position of G193, mutated to R in the Saccharomyces cerevisiae rev1-1 allele is indicated. The amino acids deleted in our REV1 BRCT mutant are shown in light type. B. Schematic for gene targeting. A 6.059 kb genomic BstBI fragment containing exons 2 -4 was amplified with primers BRCT5F1 and RBRCT3R2. A 792 bp SpeI fragment containing exon 3 was removed. It was replaced with an modified SpeI fragment in which amino acids 69 -116 of exon 3 had been removed . This fragment was generated by amplifying from the 5' SpeI site to amino 68 (2BRCTF1 and 2BRCTR1) and from amino acid 117 to the 3' SpeI site (2BRCTF2 and 2BRCTR2). The two products were joined by an ApaI site. A BamHI linker was cloned in the SalI site in the intron between exons 3 and 4 to allow introduction of the antibiotic selection cassette. The resulting cDNA is shown in pink with the position of the PCR primers used in panel D indicated. C. Southern blot showing second allele targeting. The first allele was targeted with a construct containing a blasticidin resistance cassette (bsr). The second construct contained a puromycin-resistance cassette (puro). Clones 2 and 5 have not targeted in the second round, while clone 6 has undergone retargeting of the initially targeted allele. Clones 3 and 4 have undergone targeting of both alleles. D. RT-PCR to confirm disruption of exon 3. Primer pairs are indicated on the cDNA map in panel B and detailed, along with the other primers used, in Supp. Table 2. The crossover in clone 4 has resulted in integration of the puro resistance cassette but no alteration of exon 3. Clone 3 is correct. The selection cassettes were removed by transient expression of a plasmid expressing Cre recombinase, which targets the indicated loxP sites. E. Colony survival following exposure to methyl methane sulphonate (MMS). Wild type = boxes, rev1 = circles and rev1[∆BRCT] clone #3 = inverted triangles. A. Replication stalled at a G4 structure that has formed on the leading strand template between the replicative helicase and polymerase. B. A specialised helicase (blue circles) starts to unwind the G4 exposing the first run of Gs. The deoxycytidyl transferase activity of REV1 may facilitate stabilisation of these exposed dG bases by extending the primer terminus with dC. This prevents the G4 refolding since three G runs does not support G quadruplex formation (Fig. 2B). The polymerase-interaction domain in the extreme C terminus of REV1 then coordinates the completion of the replication of the G4 DNA.  Table S2.  Table S1 is an Excel workbook with two sheetsone for the identified induced genes and the second for the control set. mRNA from three clones of wild type and rev1 were analysed. The raw expression data is given in columns E to J. The position of potential G4 sequences is given in column R relative to the transcriptional start site. nd = not done; this was because either the probe was not annotated, had multiple hits in the genome or the available genomic sequence had gaps. The sequence of the potential G4 is given 5' -3' in column S and the strand, either feature (F) or template (T) in column T. The 'Induced set' sheet lists all the genes whose basal expression in wild type is < 7 log units and which are > 1.4-fold higher in rev1 cells. Column U presents the analysis of the number of G4 DNA-containing sequences for all the genes meeting these criteria. Column V only includes those with a t-test p value of < 0.075. The control set of genes were selected at random from a pool of genes expressed at < 7 log units in wild type cells and whose expression was unchanged in the absence of REV1. The figure of 38% of these genes containing a potential G4 DNA is very similar to the figure obtained in previous analyses of G4 sequences near promoters in the chicken (Du et al., 2007) and human genomes (Eddy and Maizels, 2008;Huppert and Balasubramanian, 2007).