Haploinsufficiency of an RB–E2F1–Condensin II Complex Leads to Aberrant Replication and Aneuploidy

Fidelity of DNA replication and cell division are critical processes in multicellular organisms. Unrepaired errors can be passed on to daughter cells and contribute to cancer (1). In general, damaged DNA signals the cell cycle to arrest, and repair is undertaken before advancement into mitosis (2). Recent evidence suggests exceptions to this rule, as damage observed before mitosis is transmitted through M-phase (3, 4). These DNA lesions are often associated with replication stress (4, 5), and their ability to evade checkpoints suggests that their impact on genome instability and cancer may be significant (6).

Replication stress can lead to fork stalling and chromosomal aberrations (7). It can create short gaps in sequence, or unresolved replication intermediates (3, 7, 8). Much of the evidence for replication stress phenotypes has been demonstrated using repetitive elements in yeast that require the Condensin complex for fidelity of replication and accurate segregation in mitosis (7, 9, 10). Mammals contain two Condensin complexes, but only Condensin II is constitutively nuclear, suggesting it may play roles in both DNA replication and chromosome condensation (11). In addition to its role in mitotic chromosome condensation, chromosome shape, and controlling recombination, Condensin II has also been implicated in DNA replication and damage repair (12). In particular, Condensin II functions in the resolution of sister chromatids immediately following replication in S-phase (13). Despite these roles for Condensin II, we know little about how it is recruited to specific genomic locations, such as repetitive sequences, to carry out these functions.

The retinoblastoma protein (pRB) is generally thought of as a regulator of the G₁ to S-phase transition (14). However, evidence suggests that pRB also contributes to functions beyond G₁. For example, pRB has been implicated in an S-phase checkpoint to repair DNA breaks and to regulate initiation of DNA replication (15–18). In addition, pRB facilitates chromosome condensation in mitosis, particularly at pericentromeres, although there is no evidence that it physically localizes to this region (19–22). pRB-deficient cells have increased genome instability characterized by spontaneous DNA breaks and chromosome missegregation in mitosis (23). Surprisingly, these outcomes have not been attributed to a specific mechanism of pRB action. For example, altered regulation of E2F transcription has been implicated in shifts in intermediary metabolism (24, 25), misexpression of spindle assembly checkpoint genes, and imbalances in nucleotide pools (23). Thus, an in-depth understanding of pRB function in genome stability has yet to emerge, in part because these phenotypes are not mutually exclusive.

In humans, the RB1 gene is lost in hereditary retinoblastoma, and survivors face an elevated risk of other cancers, such as osteosarcomas, throughout their lives (26). RB1 loss occurs by the “two hit” model proposed by Knudson (27). Because loss of heterozygosity was the rate-limiting step in the genesis of retinoblastoma, Knudson concluded that heterozygosity likely did not contribute to tumorigenesis. This premise is recapitulated in Rb1^+/− mice that develop pituitary tumors characterized by loss of the remaining wild-type allele (28). Curiously, a number of experiments have suggested that loss of one copy of pRB may contribute to cancer progression in other contexts. Crosses between Rb1- and Trp53-deficient mice revealed that Rb1^+/−;Trp53^−/− mice develop some tumors without losing the remaining Rb1 allele (29). In addition, loss of one copy of Rb1 in mouse osteoblasts or embryonic stem (ES) cells confers genome instability (30, 31). Consequently, phenotypes of Rb1 heterozygosity have been previously reported, but a clear link to cancer incidence or progression is lacking. In addition, a mechanism that preserves genome stability in which the relative supply of pRB is critical has yet to be described.

pRB and Condensin II contribute to the integrity of pericentromeric heterochromatin, which is prone to the effects of replication stress. In this study, we demonstrate that Rb1-mutant cells experience replication stress, particularly at pericentromeric regions. We show that a complex composed of pRB, E2F1, and Condensin II localizes to pericentromeric repeats. Intriguingly, replication and chromosome structure defects were recapitulated in heterozygous Rb1-mutant fibroblasts, indicating that this genomic instability phenotype is gene-dosage dependent. Similarly, we demonstrate that normal RB1^+/− fibroblasts from patients with hereditary retinoblastoma exhibit the same aberrant replication characteristics followed by mitotic errors. Using genotype and copy-number variation data from the Catalogue of Somatic Mutations in Cancer (COSMIC) database, we demonstrate that RB1^+/− lymphoma and sarcoma cancer cell lines exhibit as much genomic instability as RB1^−/− cell lines. Finally, using gene-targeted mice bearing a single mutant allele that is defective for recruiting Condensin II to chromatin (called Rb1^ΔL), we demonstrate that tumors from Rb1^ΔL/+;Trp53^−/− mice have increased chromosome copy-number variation compared with Trp53^−/− controls. This provides proof-of-principle that dosage sensitivity of a pRB–E2F1–Condensin II complex compromises replication fidelity and leads to aneuploidy.

In normal human fibroblasts, loss of RB results in γH2AX foci (32). We compared γH2AX foci levels in mouse embryonic fibroblasts (MEF) from Rb1^+/+ and Rb1^−/− animals (Fig. 1A). In addition, we analyzed a targeted mutant (Rb1^ΔL) that antagonizes interactions between Condensin II and the LXCXE binding cleft region on pRB, while leaving E2F interactions and many pRB-dependent aspects of cell-cycle control intact (19, 20, 33, 34). There was a significant increase in γH2AX foci in Rb1^−/− MEFs compared with wild-type (Fig. 1B). DAPI counterstaining produces punctate foci at pericentromeric regions. Analysis by confocal microscopy revealed that γH2AX accumulation was more frequent at DAPI-rich spots in Rb1^−/− MEFs than in controls (Fig. 1C). A similar, less pronounced effect was observed in Rb1^ΔL/ΔL cells with respect to γH2AX focus abundance and localization (Fig. 1B and C). This indicates that deficiency in pRB–LXCXE–dependent functions alone is sufficient to cause γH2AX foci.

To further investigate the localization of γH2AX deposition, we performed chromatin immunoprecipitation (ChIP)-sequence analysis in which γH2AX was precipitated from wild-type, Rb1^ΔL/ΔL, and Rb1^−/− chromatin. Sequence reads were mapped onto unique and repetitive genomic regions to compare their relative abundance. Analysis on a megabase scale did not reveal a particular bias of γH2AX enrichment within damage-susceptible fragile regions (Fig. 1D). Model-based Analysis of ChIP-Seq (MACS) fails to detect reproducible enrichment in Rb1^ΔL/ΔL and Rb1^−/− cells at this location (Supplementary Fig. S1). In accordance with elevated quantities of γH2AX foci in Rb1^ΔL/ΔL and Rb1^−/− MEFs, failure to detect local regions of enrichment compared with wild-type implies that γH2AX levels are uniformly increased across unique regions of the genome in Rb1^ΔL/ΔL and Rb1^−/− cells.

Mapping reads to nonunique regions of the genome revealed that major satellite repeats, which comprise much of the DAPI-rich pericentromeric foci, were enriched relative to other nonunique comparators (Fig. 1E). ChIP-qPCR analysis also demonstrated that γH2AX is elevated at major satellites in Rb1^ΔL/ΔL and Rb1^−/− genotypes compared with wild-type (Fig. 1F and G). Therefore, deficiency for pRB increases γH2AX deposition throughout the genome, with particular enrichment at pericentromeric repeats, and loss of pRB–LXCXE interactions alone contributes to this effect.

The spontaneous formation of γH2AX foci in Rb1-mutant cells, and their frequent association with major satellite repeats, is inconsistent with DNA double-strand breaks (35). Because Rb1^−/− cells have multiple confounding defects that contribute to genome instability, we focused on the Rb1^ΔL/ΔL mutant as it maintains normal G₁–S regulation during proliferation (33).

We pulse-labeled Rb1^ΔL/ΔL cells with CIdUrd and IdUrd and stained DNA fibers to visualize replication fork progression (Fig. 2A). This revealed a net increase in fork rate in Rb1^ΔL/ΔL mutants compared with controls (Fig. 2B). Accelerated forks, although less common than slowed forks, are also associated with replication stress and are prone to stalling (5, 36, 37). In addition, elevated replication fork asymmetry (Fig. 2C), and elevated levels of replication protein A (RPA) 32 phospho-serine33, further suggest replication stress (Fig. 2D and Supplementary Fig. S2). Thus, from a genome-wide perspective, Rb1^ΔL/ΔL cells replicate their DNA abnormally and display markers of replication stress.

Because γH2AX shows greater accumulation at major satellite repeats, we investigated this region further. We used ChIP-qPCR assays of RPA to investigate its levels at major satellites in Rb1^ΔL/ΔL MEFs (Fig. 2E). This revealed lower levels of RPA at pericentromeres in Rb1^ΔL/ΔL cells, suggesting that the generation of ssDNA needed for replication or break repair is inhibited. Furthermore, ChIP-pPCR assays reveal increased levels of proliferating cell nuclear antigen (PCNA), minichromosome maintenance complex (MCM) 3, and DNA Pold at pericentromeres in Rb1^ΔL/ΔL cells (Fig. 2F–H). Together these data indicate that replication of this genomic region is defective. Investigation of γH2AX staining of mitotic cells suggests similarities between this replication phenotype and reports of under- replication (Supplementary Fig. S2). Furthermore, telomere analysis suggests no loss of integrity (Supplementary Fig. S3).

In sum, Rb1^ΔL/ΔL cells have abnormal replication rates and molecular markers that are consistent with replication stress. Although not entirely consistent with phenotypes generated by treating cells with hydroxyurea, reduced RPA levels at pericentromeres and elevated levels of replication machinery and γH2AX further suggest that replication in this region is impaired. The simplest explanation for γH2AX deposition at major satellite repeats is that they are the result of replication stress.

A similar phenotype of spontaneous γH2AX foci has been reported for a number of known pRB-interacting proteins with roles in replication. Increased γH2AX has been reported in cells deficient for the Condensin II subunit structural maintenance of chromosome protein 2 (SMC2; 38) and E2F1 (39). Notably, loss of other activator E2Fs did not cause γH2AX foci (39). In MEFs, shRNA depletion of CAP-D3 (a Condensin II subunit) caused similar patterns of γH2AX foci as described above (Supplementary Fig. S4). We also discovered similar γH2AX staining in E2f1^−/− cells (Supplementary Fig. S4).

Because of these similarities, we investigated whether pRB, E2F1, and CAP-D3 localize to pericentromeric heterochromatin in interphase cells. ChIP-qPCR for major satellites in chromatin precipitated with pRB, CAP-D3, and E2F1 antibodies revealed that they are all present at these repeats (Fig. 3A and B). Furthermore, CAP-D3 recruitment was dependent on E2F1 and pRB because its binding was diminished in E2f1^−/−, Rb1^−/−, and Rb1^ΔL/ΔL MEFs (Fig. 3B and C). To better understand Condensin II recruitment to chromatin and its relationship to pRB, we performed ChIP-seq for CAP-D3 in wild-type and Rb1^ΔL/ΔL MEFs. This revealed 8,056 peaks of local enrichment for CAP-D3 within unique genome sequences, and 6,979 of these were lost in the Rb1^ΔL/ΔL cells. Intriguingly, CAP-D3 was not recruited to the promoters of well-known pRB–E2F transcriptional targets such as Cyclin E1 (Ccne1), Mcm3, and p107 (Rbl1; Fig. 3D). However, it was recruited to the Lmnb2 replication origin in a pRB-dependent manner (Fig. 3D). This is significant because pRB and E2F1 are known to localize to this origin (16, 18), even though it does not possess a canonical E2F DNA sequence element.

We isolated the putative pRB–E2F1–Condensin II complex to investigate its properties. We generated a biotinylated major satellite repeat probe, and mixed it with nuclear extract from wild-type and Rb1^−/− cells. Major satellite-associated complexes were collected on streptavidin beads, and proteins were detected by immunoblotting. SMC2, CAP-D3, pRB, and E2F1 were isolated together from wild-type extracts, but none of these components interacted with the major satellite sequence when pRB was absent (Fig. 4A). Major satellites were bound in a sequence-dependent manner as a control probe failed to precipitate them, and their interaction was sensitive to competition by exogenous major satellite sequences (Fig. 4B). We also reconstituted this complex using nuclear extracts from human C33A cells that express HA-E2F1/HA-DP1 (Fig. 4C). Major satellite probes were mixed with extract that contained GST-RB Large Pocket (LP). Figure 4D demonstrates that neither E2F1 nor CAP-D3 was recruited to the major satellite probe without GST-RBLP. This suggests that E2F1′s ability to interact with these sequences is not autonomous, as it is in the recognition of cell-cycle target genes, but requires a cooperative contribution from pRB. Furthermore, ChIP of E2F1 from Rb1^−/− MEFs revealed cooperativity in vivo as E2F1 binding was diminished in the absence of pRB (Fig. 4E).

Previous reports indicate that binding of E2F1 to pRB through an alternate configuration called the “specific” interaction changes the sequence specificity of E2F1, reducing affinity for canonical E2F recognition sequences (40). We used MEFs from two new Rb1-mutant mouse lines to investigate whether the “specific” interaction explains cooperative DNA binding at major satellites by pRB and E2F1. Rb1^ΔG contains two amino acid substitutions (R461E and K542E) that prevent binding to activator E2Fs to regulate transcription at cell-cycle promoters; however, it maintains the ability to bind to E2F1 through the separate “specific” mechanism (41, 42). Conversely, Rb1^ΔS is unable to bind E2F1 through the “specific” interaction because of an F832A substitution, but it maintains its ability to bind activator E2Fs in a manner that regulates canonical E2F transcription (42, 43). ChIP for CAP-D3 in Rb1^ΔG/ΔG MEFs demonstrated no change in its loading at pericentromeric heterochromatin, whereas in Rb1^ΔS/ΔS cells, CAP-D3 was absent (Fig. 4F).

This indicates that pRB, E2F1, and Condensin II form a stable complex with major satellite repeats, and other unique loci in the genome such as the Lmnb2 origin (Supplementary Fig. S5). On the basis of phenotypic similarities of γH2AX focus location, we suggest that these proteins form a complex to facilitate DNA replication and chromosome condensation, and its absence preferentially affects major satellite repeats.

Major satellites account for 3% of the mouse genome, suggesting that gene dosage of Rb1 may be important for supplying enough of this complex to ensure stability of these repeats. Fluorescence microscopy revealed a significant increase in γH2AX foci in both Rb1^+/− and Rb1^ΔL/+ MEFs compared with the wild-type control (Fig. 5A and B). Moreover, γH2AX foci were found at pericentromeric heterochromatin in Rb1^+/− and Rb1^ΔL/+ MEFs more frequently than in Rb1^+/+ controls, and were almost as abundant as their homozygous mutant counterparts (Fig. 5C). Western blotting revealed elevated γH2AX levels in Rb1^ΔL/ΔL, Rb1^+/−, and Rb1^−/− MEFs relative to Rb1^+/+ (Fig. 5D). Similarly, phosphorylated RPA-S33 displays increased abundance in all Rb1-mutant genotypes (Fig. 5E). As with Rb1^−/− and Rb1^ΔL/ΔL, ChIP-seq analysis of the distribution of γH2AX in Rb1^ΔL/+ MEFs also revealed enrichment at major satellites (Fig. 5F). Using ChIP-seq data from all three Rb1-mutant genotypes, we compared the reproducibility of enrichment among repeat elements (Supplementary Fig. S6). This revealed that major and minor satellites have a greater proportion of reads than wild-type. This is noteworthy because minor satellites are adjacent to major satellites at the centromere. To extend our investigation of Rb1 gene dosage sensitivity, we performed ChIP-qPCR for CAP-D3 at pericentromeric heterochromatin in Rb1^ΔL/+ cells and determined that it was reduced at major satellites in Rb1^ΔL/+ MEFs relative to wild-type control (Fig. 5G). This suggests that loss of LXCXE interactions in even one copy of Rb1 reduced the supply of this complex at these repeats. In addition, heterozygous Rb1-mutant MEFs displayed anaphase bridges, tangled centromeres, and aneuploidy as reported previously for homozygous Rb1 mutants (Supplementary Fig. S7 and S8).

These data demonstrate that single Rb1 null and Rb1^ΔL alleles compromise the ability to prevent γH2AX foci and aneuploidy. γH2AX and phosphorylated RPA-S33 levels are similar between heterozygous and homozygous genotypes, and these translate into a similar frequency of mitotic errors. On the basis of these direct comparisons of homozygous mutant, heterozygous, and wild-type genotypes, we describe Rb1 as haploinsufficient for this function.

The kinetics of RB1 loss in retinoblastoma gave rise to the now famous “two-hit” hypothesis (27). However, our data on haploinsufficiency in Rb1-mutant mouse fibroblasts motivated us to determine whether a similar phenotype is present in normal fibroblasts from patients with hereditary retinoblastoma (RB1^+/−). We obtained patient fibroblasts, confirmed their heterozygous status by sequencing (Supplementary Fig. S9), and compared them with unrelated human fibroblast cells (IMR90, BJ, and WI38). This revealed that patient fibroblasts exhibited increased γH2AX foci (Fig. 6A and B). As with heterozygous mouse fibroblasts, they also contained elevated levels of phosphorylated RPA-S33 compared with control (Fig. 6C). We also used video microscopy of H2BGFP-expressing cells to search for mitotic defects (Fig. 6D). All three patient fibroblast lines exhibited a significant delay in progression to metaphase from the onset of chromosome condensation (Fig. 6E). Finally, patient cells also showed a significant increase in mitoses with anaphase bridges (Fig. 6E). Taken together, these experiments suggest that normal human fibroblast cells heterozygous for RB1 are characterized by replication and mitotic errors similar to our mouse data.

Given the direct impact of haploinsufficiency for RB1 on γH2AX abundance and mitotic errors, we sought evidence for aneuploidy in RB1^+/− cancers. Using copy-number variation data from COSMIC (44), we asked whether cancer-derived cell lines heterozygous for RB1 exhibit increased levels of genomic abnormalities. Because retinoblastoma survivors are highly prone to second primary neoplasms that are of mesenchymal origin (26), we compared copy-number data for RB1^+/+, RB1^+/−, and RB1^−/− cell lines from this germ layer. We sorted genomic abnormalities into four categories: whole-chromosome gains and losses, total genomic segments, total abnormal genomic segments, and chromothriptic regions. RB1^+/− and RB1^−/− cancers exhibited significantly more whole-chromosome changes than cell lines retaining both wild-type copies of RB1 (Fig. 6F). Importantly, RB1^+/− lines exhibited as many whole-chromosome changes as RB1^−/− cells. This trend was observed for the other measures of instability as RB1^+/− and RB1^−/− lines exhibited significantly more abnormalities than wild-type and were not statistically different from each other (Fig. 6G–I). To confirm these results, we obtained representative lines to verify RB1 copy number and expression of pRB (Supplementary Fig. S9).

These data indicate that loss of one copy of RB1 may exhibit haploinsufficiency in humans, namely through the ability of pRB to prevent the accumulation of γH2AX foci, phosphorylated RPA-S33, and mitotic errors. Moreover, this is associated with increased chromosomal abnormalities in cell lines at a level comparable with that found in RB1^−/− cells. A number of possibilities exist that may connect replication and chromatin structure defects at pericentromeric sequences with common chromosomal aberrations found in cancer, and they are discussed below.

We also sought evidence for haploinsufficiency in vivo using mouse models. Because Rb1^−/− mice are inviable, comparisons between heterozygous, homozygous mutant, and wild-type populations are not possible. However, Rb1^ΔL/ΔL mice are viable and can therefore be used to study whether haploinsufficiency in Rb1^ΔL/+ may contribute to oncogenesis. We crossed Rb1^ΔL/+ mice with Trp53^−/− mice as previously reported (19). Rb1^ΔL/+;Trp53^−/− mice exhibited a significant decrease in survival compared with Trp53^−/− controls (Fig. 7A), which was not statistically different from previously reported Rb1^ΔL/ΔL;Trp53^−/− mice (19). PCR amplification of the wild-type and Rb1^ΔL alleles in tumor DNA from Rb1^ΔL/+;Trp53^−/− mice showed retention of the wild-type allele (Fig. 7B). Thymic lymphomas from Rb1^ΔL/+;Trp53^−/− mice demonstrated a more aggressive histology compared with Trp53^−/− controls as evidenced by smaller cytoplasmic-to-nuclear area (Fig. 7C). This is very similar to the histology described in Rb1^ΔL/ΔL;Trp53^−/− thymic lymphomas (19), and Rb1^ΔL/+;Trp53^−/− mice exhibited a shift in tumor spectrum from thymic lymphoma, commonly reported in Trp53^−/− mice, toward sarcomas (Supplementary Fig. S10). Thus, one defective copy of Rb1 can have similar effects on disease progression as mutation of both alleles. Because thymic lymphomas were common among all genotypes, we investigated their relative levels of chromosome gains and losses. Figure 7D shows log₂ ratio plots from a male versus female control hybridization and representative Rb1^ΔL/+;Trp53^−/− tumors. The number of whole-autosome gains and losses in thymic lymphomas from Rb1^ΔL/+;Trp53^−/− mice was greater than Trp53^−/− controls, and statistically indistinguishable from Rb1^ΔL/ΔL;Trp53^−/− tumors (Fig. 7E).

Earlier experiments demonstrated that one Rb1^ΔL allele can cause replication and chromosome structure–associated phenotypes. Taken together with this tumor study, our data suggest that gene dosage-sensitive effects on replication and mitotic chromosomes have the potential to compromise pRB function in vivo leading to aneuploidy.

The pRB–E2F1–Condensin II complex identified in this study affects chromatin from S-phase to mitosis. Our findings suggest how replication and structural defects at pericentromeres may be connected to common chromosomal abnormalities in cancer. Reduced occupancy by this complex at pericentromeric repeats may lead to unresolved replication intermediates and compromise the integrity of centromeres and kinetochores. Misshapen kinetochores can lead to merotelic microtubule attachments, lagging anaphase chromosomes, and ultimately gains and losses of whole chromosomes in daughter cells (45). Alternatively, centromere fusions or recombination with other chromosomes (7, 8) may lead to gains or losses of whole chromosome arms, and these are the most common segmental chromosome changes in human cancer (46). Finally, lagging chromosomes created by these mechanisms are more likely to become incorporated into micronuclei and undergo chromothripsis in subsequent cell cycles (47). In this way, defects in pericentromeric chromosomal regions may have the ability to cause a wide spectrum of chromosomal abnormalities, similar to those found in RB1^+/− and RB1^−/− cancer cell lines. Intriguingly, augmentation of Cohesin function can suppress replication and chromosomal abnormalities caused by loss of RB1 function. This suggests that these chromosomal defects may potentially be suppressed therapeutically (48).

We were surprised to discover that pRB's function with E2F1 and Condensin II was sensitive to gene dosage. Knudson's hypothesis, that both copies of RB1 are lost during tumorigenesis (27), has greatly shaped our genetic understanding of cancer. In humans, hereditary retinoblastoma survivors acquire early-onset second primary cancers more frequently than the general population (26). Thus, it is interesting to speculate that the underlying genomic instability we observe in RB1^+/− patient fibroblasts could contribute to this increased penetrance. The secondary tumors in retinoblastoma survivors are often sarcomas (26,49), which is consistent with the genome instability characteristics of mesenchymal cancers highlighted in this study.

Our data bring together concepts of replication and mitosis and demonstrate how abundance of a pRB–E2F1–Condensin II complex may link them. RB1^+/− patient fibroblasts are haploinsufficient in their ability to prevent the accumulation of γH2AX and mitotic errors, suggesting that RB1 haploinsufficiency in this functional context may contribute to tumorigenesis in humans.

Primary MEFs were prepared and cultured according to standard methods. Rb1^ΔG/ΔG and Rb1^ΔS/ΔS introduce R461E/K542E and F832A coding substitutions into the murine Rb1 gene, respectively (41, 42). The generation of Rb1^ΔS/ΔS mice will be published elsewhere. Mitotic chromosome spreads were prepared and H2BGFP transduction of MEFs was as before (19). For DNA combing, cells were pulse-labeled for 30 minutes with CIdUrd, followed by 30 minutes of IdUrd as described previously (50).

Primary patient fibroblasts were obtained from the Coriell Institute for Medical Research (Camden, NJ) and cultured as recommended. RB1 genotyping was performed by Impact Genetics Inc. We created a FUtdTW lentiviral vector that expresses H2BGFP for analysis of human fibroblasts. All cell lines and their culture conditions are described in the Supplementary Methods.

Stained cells were examined on an Olympus Fluoview FV1000 confocal microscope system, and z stacks at intervals of 7 μm were collected using the Olympus Fluoview FV1000 Viewer. Collapsed and 3D rendered images were used to determine whether colocalization of γH2AX foci coincided with pericentromeric DNA.

Live-cell microscopy was carried out as described previously (19). FISH analysis of aneuploidy and telomeres was carried out using previously reported methods (33). CIdUrd and IdUrd staining of DNA fibers was carried out as described previously (50) with modifications described in the Supplementary Methods. Fork velocity was calculated as IdUrd length/time. Percentage asymmetry is [(long IdUrd track - short IdUrd track) - 1] × 100.

Cycling cells were fixed in 1% formaldehyde, except ChIP experiments to detect PCNA, MCM3, and DNA Polδ that were fixed in 1% formaldehyde and ethylene glycol bis[succinimidylsuccinate] (EGS). ChIP methods and the source of PCR primers are described in the Supplementary Methods. For ChIP-seq experiments, γH2AX and CAP-D3 antibodies were used to precipitate chromatin, and 150 ng of DNA was used for library preparation and sequenced on an Illumina Hi-Seq 2000 at The Centre for Applied Genomics (Sick Kids, Toronto, ON, Canada). Unique sequence reads were aligned to the mm9 genome assembly or repeat containing indexes and analyzed as described in the Supplementary Methods. Raw sequence data are available in the Gene Expression Omnibus (GEO; GSE55041).

Nuclear extracts from the indicated cells were prepared as described previously (42). For isolation of pRB–E2F1–Condensin II complexes, a major satellite repeat probe was generated by PCR using a biotinylated primer. Extracts were mixed with the probe and complexes were precipitated using streptavidin dynabeads. Western blotting identified captured proteins. GST pull-downs were carried out by standard methods. Antibodies used in this study can be found in the Supplementary Methods.

The Rb1^ΔL-mutant strain (Rb1^tm1Fad) has been previously described (33). The Trp53^−/− mice were obtained from The Jackson Laboratory. All animals were housed, handled, and analyzed as previously described (19). Tissues were processed for staining according to standard methods. Array comparative genomic hybridization (CGH) data were processed as previously reported (19), and are available in GEO (GSE51876). A list of cell line data analyzed from COSMIC is present in the Supplementary Methods.

No potential conflicts of interest were disclosed.

Conception and design: C.H. Coschi, F.A. Dick

Development of methodology: C.H. Coschi, C.A. Ishak, A. Marshall, S. Talluri, J. Wang, V. Percy, I. Welch, P.C. Boutros, F.A. Dick

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.A. Ishak, D. Gallo, S. Talluri, M.J. Cecchini, A.L. Martens, V. Percy, I. Welch

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.H. Coschi, C.A. Ishak, D. Gallo, A. Marshall, S. Talluri, J. Wang, I. Welch, P.C. Boutros, G.W. Brown, F.A. Dick

Writing, review, and/or revision of the manuscript: C.H. Coschi, C.A. Ishak, P.C. Boutros, G.W. Brown, F.A. Dick

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.A. Ishak, A. Marshall

Study supervision: P.C. Boutros, F.A. Dick

The authors thank Nathalie Bérubé and Kristin Kernohan for discussions and advice.

C.H. Coschi was a member of the Cancer Research and Technology Transfer (CaRTT) training program and was a Canadian Institutes of Health Research (CIHR) doctoral award recipient. D. Gallo was supported by an NSERC (PGS-D) award. A. Marshall holds a Natural Sciences and Engineering Research Council (NSERC) CGS-M fellowship. C.A. Ishak, S. Talluri, and M.J. Cecchini are also members of CaRTT, and M.J. Cecchini was supported by a CIHR MD/PHD award. F.A. Dick is the Wolfe Senior Fellow in Tumor Suppressor Genes. This study was conducted with support from the Ontario Institute for Cancer Research to P.C. Boutros through funding from the Government of Ontario and the CIHR to G.W. Brown (MOP79368) and F.A. Dick (MOP64253 and MOP89765).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.