Rad21-Cohesin Haploinsufficiency Impedes DNA Repair and Enhances Gastrointestinal Radiosensitivity in Mice

Approximately half of cancer-affected patients receive radiotherapy (RT). The doses delivered have been determined upon empirical experience based upon average radiation responses. Ideally higher curative radiation doses might be employed in patients with genuinely normal radiation responses and importantly radiation hypersensitive patients would be spared the consequences of excessive tissue damage if they were indentified before treatment. Rad21 is an integral subunit of the cohesin complex, which regulates chromosome segregation and DNA damage responses in eukaryotes. We show here, by targeted inactivation of this key cohesin component in mice, that Rad21 is a DNA-damage response gene that markedly affects animal and cell survival. Biallelic deletion of Rad21 results in early embryonic death. Rad21 heterozygous mutant cells are defective in homologous recombination (HR)-mediated gene targeting and sister chromatid exchanges. Rad21+/− animals exhibited sensitivity considerably greater than control littermates when challenged with whole body irradiation (WBI). Importantly, Rad21+/− animals are significantly more sensitive to WBI than Atm heterozygous mutant mice. Since supralethal WBI of mammals most typically leads to death via damage to the gastrointestinal tract (GIT) or the haematopoietic system, we determined the functional status of these organs in the irradiated animals. We found evidence for GIT hypersensitivity of the Rad21 mutants and impaired bone marrow stem cell clonogenic regeneration. These data indicate that Rad21 gene dosage is critical for the ionising radiation (IR) response. Rad21 mutant mice thus represent a new mammalian model for understanding the molecular basis of irradiation effects on normal tissues and have important implications in the understanding of acute radiation toxicity in normal tissues.


Introduction
Radiotherapy (RT) is employed in approximately half of all cancer patients. Most RT is delivered to a cancer-affected organ, unavoidably encapsulating normal tissues surrounding a tumor. Optimal outcomes are a balance between normal tissue toxicity and tumour control: this balance is known as the therapeutic ratio. In some cases, RT is delivered with very large fields, such as for total/whole body irradiation (T/WBI), which is used in hematopoietic stem cell transplantation in patients with hematologic and other malignancies.
Although some rare syndromes are characterised by radiosensitivity, mechanistic insights of the effects of IR on the majority of patients' normal tissues and tumours are largely lacking. For example rare (1 in 100,000-200,000 newborns) Ataxia telangiectasia (Atm) patients are hypersensitive to RT [1][2][3][4]. While Atm heterozygous carriers are relatively common (1 in 200 individuals) and have an elevated risk for cancer particularly breast cancer [5][6][7], they are not reliably identified by hypersensitivity to RT or to IR in vitro. Identification of potential new mammalian radiation responsive genes/pathways promises refinement of treatment protocols and the possibility of tailored therapy. Nevertheless this has been an elusive goal.
Rad21 (also known as Scc1 or MCD1) was initially identified in a genetic screen in fission yeast (Saccharomyces pombe) [8]. This clone contains a point mutation in one allele of the Rad21 gene and exhibited hypersensitivity to radiation owing to its impaired double strand DNA breakage repair [8]. Recent studies in lower eukaryotes and metazoa have shown that paralog mammalian Rad21 genes are key regulatory components of a multi-protein complex, cohesin [9][10][11][12][13][14]. Cohesin plays an essential role in mediating sister chromatid cohesion (SCC), a mechanism critical for proper chromosome segregation [10][11][12]. Biallelic deletion of cohesin subunits results in cell death [10][11][12]. Recent studies implicate cohesin in the DNA damage response and repair in eukaryotic cells [13][14][15][16]. We previously found Rad21 variants in cancer patients exhibiting acute radiation toxicity, suggesting an association between Rad21 gene variants and normal tissue protection that may be defective in some radiation sensitive cancer patients [17].
To gain insight into the contribution of Rad21 cohesin to normal tissue IR toxicity, we generated mutant mice with one functional allele of Rad21 and investigated the IR response of this mutant in the context of whole animals. Our study provides the first evidence that Rad21, and possibly cohesin and their associated genes, represent a new class of novel total body radiation response gene(s) in mammals, the characterisation of which has important implications for patient-tailored cancer therapy and modulating normal tissue responses to clinical RT.

Deletion of the mouse Rad21 gene results in early embryonic lethality
To generate a null allele of the mouse Rad21 gene, we constructed a targeting vector which replaced exon 2 of the mouse Rad21 gene with a neo resistance cassette flanked by lox P sites, resulting in a null allele (Fig. S1A). The targeting construct was introduced to mouse embryonic stem (ES) cells and mutant cells were produced by homologous recombination. Targeted alleles were identified by PCR and Southern blot analysis (Fig. S1A).
Chimeric offspring were obtained from Rad21 +/2 ES cell clone 5 which showed a reduced RAD21 protein level (Fig. S1B) and heterozygous Rad21 +/2 mice were obtained by breeding. Rad21 +/2 mice were viable and developed to apparently normal adulthood without morphological defects. Genotyping of 255 live-born offspring (n = 40 litters) from heterozygous parents revealed no homozygous animals, suggesting that homozygous null Rad21 mice die in utero. Further examination of embryos at embryonic gestational times of E8.5 (n = 7), E10.5 (n = 15) and E13.5 (n = 25) days from heterozygous parents, revealed no homozygous mutants at all three developmental stages examined. Together, these data indicate that homozygous Rad21 deficiency leads to early embryonic lethality. Thus, at least one WT Rad21 allele is essential for normal development in mammals and the generation of homozygous knock-out cells and embryos for study was precluded by this traditional approach. Nevertheless, Rad21 heterozygous cells and mice were available for investigation.

Rad21 +/2 mouse embryonic fibroblasts (MEFs) exhibit increased chromosomal number abnormality and mitotic defects
Deletion of Rad21 compromise proper chromosome segregation, leading to gains or losses of chromosomes (aneuploidy) in yeast and vertebrates [10,11,14]. To determine whether chromosomal number is altered in heterozygous Rad21 knock-out cells, we examined metaphase chromosome spreads in early passage MEFs from Rad21 +/2 and isogenic WT mice. Cells with diploid chromosome numbers (i.e. 40 chromosomes) were observed in approximately 56% of WT cells (Fig. S2A). In contrast, a clear reduction in cells with diploid chromosome content was observed in Rad21 +/2 MEFs (Fig. S2A). This was accompanied by an increase in the percentage of aneuploid cells (Fig. S2A). Further examination of chromosome spreads showed that Rad21 +/2 cells often had one or more pairs of sister chromatids with no apparent connection at centromeres, suggesting the premature loss of, or weakened centromeric cohesion, or an altered centric chromatin organization (Fig. S2B). Furthermore, Rad21 +/2 MEFs showed more frequent mitotic defects compared to WT (Fig. S2C). The commonly observed mitotic defects in Rad21 +/2 MEFs include uncondensed or mis-aligned chromosomes at metaphase, chromosome bridges and lagging at anaphase, and multipolar mitosis (Fig. S2C).
Response of Rad21 +/2 cells to DNA damaging agents Both budding and fission yeast Rad21 mutants are hypersensitive to IR and defective in DNA repair [8,18]. Likewise, a conditional Rad21 deletion in chicken DT-40 cells led to an increase in spontaneous and IR-induced chromatid breaks [14]. To test whether Rad21 +/2 mouse cells are IR sensitive, we determined the clonogenic survivals of WT and Rad21 +/2 ES cells following IR. No significant difference was observed between the Rad21 +/2 and WT cells at any dose tested (Fig. 1A). We then examined the clonogenic survival of Rad21 +/2 ES cells after treatment with the DNA cross-linking agent, Mitomycin C (MMC) in view of the evidence that the predominant means of DNA repair of MMC-induced damage is mediated by HR repair [19]. In response to MMC treatment, Rad21 +/2 ES cells displayed a significantly reduced survival compared to WT (p,0.001) (Fig. 1B).
Reduced gene targeting efficiency at an independent locus, and impaired MMC-induced SCE formation in Rad21 +/2 cells reveal in vivo and in vitro HR deficiency Hypersensitivity of Rad21 +/2 cells to cross-linking agents like MMC raises the prospects of an associated defect in HR repair pathways. Certainly, SCC genes have been implicated in promoting HR [13,16]. However, it is unknown if the HR role of SCC genes is conserved in mammalian cells. We therefore assessed the HR capability of Rad21 +/2 cells using an eGFP-based assay which employs a Rad54-GFP knock-in construct targeting the genomic Rad54 locus in ES cells [20]. In this gene targeting assay, cells are transfected with the promoter-less hRad54GFPpuro knock-in targeting construct. Cells containing integrated construct (via gene targeting or random integration) were then selected in medium containing Puromycin. GFP-positive cells arise when the targeting construct integrates in the mRad54 locus via homologous recombination but not random construct integration. The frequency of these events is detected by GFP-fluorescence of targeted ES cells by flow cytometry, an indirect measure of targeting efficiencies. A Rad54 2/2 ES cell line was also employed as a gene targeting-defective control. As expected in Rad54 2/2 ES cells, there was a significant reduction in GFP-positive cells (5%) when compared to 30% GFP-positive cells in the WT isogenic cell line, 1B10 (Fig. 1C). The Rad21 +/2 ES cell line exhibited an approximately two-fold reduction in targeting efficiency (17%) when compared to its WT parental ES cells (35%), indicating a HR defect in live Rad21 +/2 ES cells. This result is in contrast to a previous report which showed an increased gene targeting efficiency in human 293 cells following a transient knockdown of Rad21 by siRNA [16]. The discrepancy may be due to a specific function of RAD21 in HR, since the previous study measured episomal recombination following the site-specific induction of a DSB by the I-SceI endonuclease [16]. This targeting measures different recombination events than the homologous recombination event we measured by targeted integration. It is also possible that the severe deficiency of Rad21 by siRNA knockdown may shift the preference to a different sub-pathway of HR.
To further probe the evidence for a role of Rad21 in HR, we examined in Rad21 mutant cells whether somatic recombination activity is also altered, by measuring the level of sister chromatid changes (SCEs) using Rad21 +/2 MEF cell lines derived from Rad21 +/2 mice. SCE has been shown to be closely associated with HR in vertebrate cells and the frequency of SCE serves as a measure of recombination activity [21]. Both WT and Rad21 +/2 mutant MEFs showed only spontaneous SCE levels of three per metaphase spread (data not shown). No significant difference between WT and Rad21 +/2 MEFs was observed. However, following treatment with 6 mM MMC, Rad21 +/2 MEFs showed a modest increase in SCE frequency with an average of five SCEs per metaphase spread (Fig. 1D). In contrast, the frequency of SCEs in WT MEFs was elevated substantially to an average of fourteen SCEs per metaphase spread, significantly higher (p ,0.0001) than that of Rad21 +/2 MEFs (Fig. 1D). The suppression of SCEs in Rad21 +/2 MEFs is similar to that observed in human cells following siRNA knockdown of Rad21 [16] and in known recombination-defective cells [21], providing further evidence for defective HR in Rad21 +/2 MEFs. Rad21 +/2 MEFs are defective in the DNA damage checkpoint activation and DSB repair Two key cellular responses to IR damage are checkpoint activation and DNA repair, the two processes being closely linked. To determine if Rad21 +/2 MEFs are defective in DNA damage-induced cell cycle arrest, we analysed cell-cycle profiles of Rad21 +/2 and isogenic WT MEF cells following IR. While similar cell cycle distributions were observed in unirradiated Rad21 +/2 and WT cells, the two cell types exhibited different profiles when examined 24 hours post IR ( Fig. 2A). Rad21 +/2 cells showed a higher percentage of S phase cells compared to WT cells following IR ( Fig. 2A). Further analysis using BrdUpulse-labelling revealed approximately 71% reduction of cells undergoing DNA replication in WT MEFs following IR (Fig. 2B). Interestingly, there was only a modest reduction (approximately 29%) in BrdU-positive cells in Rad21 +/2 MEFs, following IR (Fig. 2B). These data showed that the majority of Rad21 +/2 cells continues to undergo DNA synthesis following IR, suggesting defective intra-S checkpoint activation. In addition to S phase cells, an increase in G2/M cells was observed in Rad21 +/2 MEFs following IR ( Fig. 2A). Further examination of nuclear morphology using DAPI staining revealed that only 1% of mitotic cells were present in WT 24 hours post IR, indicating that WT cells were prevented from entering mitosis ( Fig. 2C and Fig. S3). By contrast, 18% of Rad21 +/2 cells were engaged in mitosis. Strikingly, nucleoplasmic bridges (NPB) were present in most, if not all, mitotic Rad21 +/2 cells ( Fig. 2C and Fig. S3), suggesting that a delayed progression through mitosis may contribute to NPB accumulation. Furthermore, cells with micronuclei (MN) were found to be more frequent in Rad21 +/2 MEFs (,13%) than WT (5%). By 48 hours post IR, cells with NPB and MN were increased dramatically in Rad21 +/2 MEFs to 25% and 43%, respectively ( Fig. 2C and Fig. S3). The presence of NPB and MN suggest that Rad21 +/2 cells more frequently enter mitosis with damaged DNA, resulting in both chromosomal and mitotic abnormalities.
We next examined whether IR-induced DSB repair is impaired in Rad21 +/2 cells using cH2AX as a reliable surrogate marker for DSBs [22]. Of two Rad21 +/2 MEF cell lines examined, both showed significantly higher numbers of cH2AX foci when compared to that of isogenic WT cells ( Fig. 2D and Fig. S4). These data confirm that DSB repair is defective in Rad21 +/2 MEFs. Rad21 +/2 mice are sensitive to WBI Although studies in Rad21 +/2 ES cells did not reveal apparent radiation hypersensitivity in vitro, we were mindful that in vitro survival assay may not be representative of somatic cell types in vivo. Rad21 +/2 mice are one of six thus far described cohesin mouse mutants [23][24][25][26][27], but to date these mutants have not been studied to assess the effect of cohesin deficiency on radiation response in the context of whole animals. This might be important with regard to patient responses to RT. Therefore we examined the effects of WBI on the survival of Rad21 +/2 mice.
Rad21 +/2 mutant and WT littermates were exposed to a single dose of 8 or 13Gy WBI and Kaplan-Meier survival curve was determined over 30 days post IR. Following a single dose of 8Gy WBI, one death was recorded for WT mice. By contrast, a significant reduction in post-irradiation survival was observed in Rad21 +/2 mutant mice, beginning on day 17 post IR (Fig. 3A). To examine these data in the context of radiation sensitivity, we treated a cohort of Atm-defective mice with one or two alleles harboring a kinase domain inactivation mutation found in AT patients (AtmSRID) [28]. As expected, the AtmSRID homozygous mutant mice were remarkably hypersensitive to IR, such that no animals survived beyond day 8. However, we found that AtmSRID heterozygous mutant mice were more resistant to WBI than Rad21 +/2 mice, with 90% surviving beyond 30 days. These data make a simple but very important point that Rad21 +/2 heterozygous mice are extraordinarily hypersensitive to IR, even from the perspective of the archetypical IR sensitivity of mutations in the Atm gene.
Exposure to a higher dose (13Gy) of WBI resulted in 100% mortality in both Rad21 +/2 and WT littermates by day 8 posttreatment (Fig. 3B). The death of Rad21 +/2 animals occurred as early as day 4 post-IR and ,38% Rad21 +/2 mice died by day 6, when no mortality was recorded for WT mice. The reduced survival following WBI in Rad21 +/2 mice was presumed to be due to the consequences of greater tissue damage. This was then investigated in more detail.

Rad21 +/2 mutant animals display GIT radiosensitivity
In adult mammals, the GIT represents one of the most sensitive tissues to IR [29]. Given that our immunostaining revealed that RAD21 protein is abundantly expressed in crypt epithelial cells of both small and large intestines (Fig. S5), we predicted that WBI may have greater effect on Rad21 +/2 mice compared to WT animals. We therefore performed a detailed characterization of GIT damage in Rad21 +/2 and AtmSRID homozygous mutant mice after WBI treatment. IR-induced GI crypt cell damage can be quantitatively analysed in the small intestine (SI) using the well-established in vivo microcolony assay for crypt survival [30,31]. In mouse SI, surviving crypts, defined as consisting of 10 or more cells with prominent nuclei and little cytoplasm, are readily distinguishable from nonsurviving crypts between 3 to 4 days post-IR [30,31].
We first determined the consequences of WBI on crypts at day 3.5, a time when we and others have observed maximum post-IR crypt damage, and when recovery is initiated. Following WBI, some crypts shrunk as a result of extensive cell depletion and appeared as 'ghost' structures ( Fig. 4A). Taking a more quantitative approach, we assessed the percentage of SI crypt survival at 3.5 days post WBI in Rad21 +/2 and WT littermates following graded doses of 0, 6, 8, 10 and 13Gy, using the microcolony analysis. There was no statistical difference in the number of crypts between the different genotypes under homeostasis and without IR (Table S1). Rad21 +/2 mice exhibited a dose-dependent decrease in crypt survival that is greater than WT (Fig. 4B). The difference between the genotypes was highest at 13Gy, with more than two-fold higher surviving crypts recorded in WT (57.6%) compared to Rad21 +/2 (22.7%) mice (Fig. 4B and Table S1). Interestingly, a slight increase in crypt number in WT mice following WBI at 8Gy was observed (Fig. 4B  and Table S1). Presumably, this increase in the number of crypts is associated with IR-induced crypt regeneration. These data clearly show that Rad21 +/2 mice were more susceptible to IR-induced SI damage compared to WT.
When AtmSRID homozygous mutant SI were examined, it was apparent that they were substantially more IR sensitive than WT mice (Fig. 4B). To test for synthetic interactions between Atm and Rad21 in the in vivo response of the SI to IR, we intercrossed these mice. When challenged with WBI, the high degree of SI sensitivity of the AtmSRID homozygous mutant SI was not enhanced in the presence of our Rad21 mutant allele. These data further supported the notion that, in mammals, the function of Rad21 contributes to the same pathway as Atm. The SI of mice with a single Rad21 mutant allele was demonstrably, reproducibly and statistically more IR sensitive than the AtmSRID heterozygous SI (Fig. 4C).
To assess whether the reduced crypt survival in Rad21 +/2 mice is the result of intrinsic epithelial or secondary damage (for example, as a result of endothelial cell damage), we developed an ex vivo crypt survival assay using organoids established from isolated SI crypts. In this assay, crypts were irradiated ex vivo and the organoids which formed were counted 10 days post irradiation. At an IR dose of 2.5Gy, 72% of organoids were formed from WT crypts. Only 48% organoids were formed from irradiated Rad21 +/2 crypts, significantly lower than that of WT (Fig. 4D). We noted that no significant difference was observed between WT and Rad21 +/2 crypts at a higher IR dose of 5Gy. We posit that this higher dose is above a threshold level even for WT SI, for extensive crypt death in both genotypes. Nevertheless, these data provide strong evidence that the crypt death in Rad21 +/2 mice is unlikely to be a consequence of toxicity secondary to endothelial dysfunction.
Small intestinal epithelial cells of Rad21 +/2 mice are deficient in DSB repair and fail to resume crypt regeneration following IR Given that Rad21 +/2 MEFs exhibited impaired DSB repair, we hypothesised that enhanced radiation sensitivity of Rad21 +/2 crypt cells was associated with their inability to repair IR-induced DSBs. We therefore examined the presence of cH2AX foci in the small intestinal crypts of Rad21 +/2 and WT mice 6 hours post WBI. Distinct cH2AX foci were detected in the crypts of both Rad21 +/2 and WT mice, indicating the presence of DSBs (Fig. 5A). To characterise DNA repair according to cell type, cH2AX-positive cells were scored on a positional basis from the base to the midregion of crypts, i.e. position 1 to position 10, which correspond to the combined Paneth cell, stem cell and transient amplifying compartments [32,33]. Rad21 +/2 mice showed significantly higher frequencies of crypt epithelial cells with cH2AX foci at all positions examined, when compared with WT (p,0.0001) (Fig. 5A). The persistence of cH2AX foci in high proportions of Rad21 +/2 crypt epithelial cells strongly suggests a deficiency in DSB repair and is consistent with the defect we showed in Rad21 mutant cells.
Radiation induced cell death is thought to be a major factor contributing to subsequent crypt loss [34,35]. We next examined the effect of WBI on crypt cell apoptosis. TUNEL staining of the SI at 6 hours, 24 hours and day 3.5 post WBI revealed that IR induced a significant increase in crypt cell apoptosis in both Rad21 +/2 and WT mice (Fig. 5B, left panel). The number of apoptotic cells per crypt (Apoptotic Index) was highest at 6 hours post IR and decreased sharply at 24 hours (Fig. 5B, left panel). By day 3.5, few apoptotic cells per crypt were observed. The apoptotic index was not significantly different between Rad21 +/2 and WT mice; although there is a moderately higher apoptotic index in Rad21 +/2 mice compared to WT at 24 hours and at day 3.5 post IR (Fig. 5B, left panel). We further determined the frequency of TUNEL-positive cells according to the cell position in crypts at 6 hours post WBI when the peak of apoptosis was observed. TUNEL-positive cells in Rad21 +/2 mice were found to be more frequent in the crypt base at cell positions 1 to 4 where cryptal basal cells (stem/progenitor-cells) and Paneth cells reside [35,36] (Fig. 5B, right panel), suggesting that stem/progenitor cells in Rad21 +/2 are more susceptible to IR-induced cell death.  Table S1. Error bars = SEM. C. Comparison of crypt survival in Rad21 and Atm mutant animals 3.5 days post 13Gy WBI. WT and Rad21 +/2 data shown areas in B. The number of animals used is as follows: Rad21 +/2 : n = 7; AtmSRID +/2 : n = 5; WT: n = 7. Error bars = SEM. D. Ex vivo crypt organoid survival. Diagrammatic presentation of the assay is shown above the graph. Day 1 organoid nuclei were visualised with Hoechst 33258 to show live cells. Data represent the percentage of organoids formed 10 days post IR at the doses indicated relative to unirradiated controls. The number of animals used is as follows: WT: n = 5; Rad21 +/2 : n = 7; four replicates were counted for each animal. Error bars = SEM. doi:10.1371/journal.pone.0012112.g004 The increased frequencies of cH2AXand TUNEL-positive cells at the crypt base raise the possibility that the stem cell compartment in Rad21 +/2 intestinal crypts may be more susceptible to IR damage and cell death. The recruitment of stem cells to a proliferative state has been shown to be critical for crypt regeneration and GI recovery following IR-induced damage. We therefore assessed crypt cell proliferation in the SI of Rad21 +/2 and WT mice at 3.5 day post IR, a time when crypt regeneration occurs, using PCNA as a surrogate marker of proliferation. In unirradiated mice of both genotypes, ''strings'' of PCNA-positive cells were detected in the mid-crypt region, corresponding to the transient amplifying cells (Fig. 5C). At 3.5 days post 13Gy WBI, PCNA-positive cells were present in the majority of crypts in WT mice, indicative of active cell proliferation associated with crypt regeneration (Fig. 5C). In irradiated Rad21 +/2 mice, arrays of PCNA-positive cells were detected in ''surviving'' crypts but in the majority of crypts only a few scattering PCNA-positive cells were observed (Fig. 5C). We further examined proliferation using a mitotic-specific protein marker, phospho-histone H3 (PH3) in SI crypts to ensure that the PCNA staining was not being influenced in the WT crypts by PCNA-associated DNA repair functions [37]. Consistent with our PCNA staining, abundant PH3-positive cells were detected in surviving crypts of both Rad21 +/2 and WT mice (Fig. 5D). However, the number of PH3-positive cells per crypt (Mitotic Index) increased nearly two fold in irradiated WT mice compared to unirradiated animals, suggesting active mitoses associated with crypt regeneration (Fig. 5D). In irradiated Rad21 +/2 mice, the mitotic index remained at the similar level as unirradiated animals, but was significantly lower (p,0.0001) compared to irradiated WT mice (Fig. 5D). Taken together, these results strongly suggest that defective DSB repair and increased cell death within the region of the stem/progenitor cell compartment impairs crypt regeneration, leading to enhanced SI radiosensitivity in Rad21 +/2 mice.
Colonic crypt epithelial cells of Rad21 +/2 mice fail to undergo mitotic arrest We further performed a detailed examination of the response of colonic crypt cells to IR, as radiation damage to the large intestine (colon) is similar to that of the SI but occurs at a slower rate. Colon damage doesn't form a prominent component of WBI syndromes, as its functional consequences are not as marked as for the SI. Nevertheless, in addition to its more regular crypt morphology as compared to the SI, it provides another opportunity to characterise the kinetics of crypt regeneration. We scored the frequency of PCNA-positive cells by colonic crypt cell position at day 1 and 3.5 post IR, to assess whether proliferation occurred in colonic crypt epithelial cells (Fig. 6A, 6B). The cell located in the centre of crypt base was designated as cell position 1 and cells that stained for PCNA were accordingly assigned to cell positions 1 or more up the crypt wall. In this fashion, data generated from the sides of both crypt walls can be pooled [38]. In unirradiated Rad21 +/2 and WT mice, the highest frequencies (8 to 12%) of PCNA-positive cells were detected at the positions 2 to 10; where the stem cells and lower transient amplifying population exist (Fig. 6B), with no significant difference between Rad21 +/2 and WT mice (Fig. 6B). At day 1 post IR, the frequency of PCNA-positive cells was dramatically reduced to 2-3% at these cell positions in irradiated WT colonic crypts (Fig. 6B). By contrast, the frequency of PCNA-positive cells at these positions in Rad21 +/2 mice remained similar to, or slightly increased following IR (Fig. 6B). Strikingly, the frequencies of PCNA-positive cells in irradiated Rad21 +/2 animals were significantly higher at all cell positions compared to that of irradiated WT (Fig. 6B). This response in Rad21 +/2 mice was unexpected and indicated that characteristic G1/S and G2/M checkpoints normally invoked by IR [39] were not being activated in the mutant GIT. Additionally, the frequencies of PCNA-positive cells at the positions 11 to 15 were noticeably increased following IR in the Rad21 +/2 crypts (Fig. 6B), consistent with an expected IR-induced recruitment of G0 cells into a proliferative state. At day 3.5 post IR, a striking reversed pattern was observed in WT animals, with a significant increase in the frequencies of PCNA-positive cells at cell positions 1 to 19. This was in sharp contrast to a significant reduction in the number of PCNApositive cells at the majority of crypt cell positions in Rad21 +/2 mice (Fig. 6B). Together, these data suggest that Rad21 +/2 colonic crypt cells failed to initiate the canonical early IR-damage response of mitotic arrest. This severely compromised the subsequent process of crypt regeneration in mutant Rad21 +/2 mice.
Rad21 deficiency leads to an enhanced IR sensitivity in bone marrow stem cell compartments Bone marrow contains stem cells, progenitor cells and their progeny, that are exquisitely radiosensitive [40]. Bone marrow suppression is the most recognized form of IR-induced damage after WBI in mammals, leading to irreversible depletion of the bone marrow stem cell reserve. To characterise IR-induced bone marrow damage in Rad21 +/2 animals, we investigated the hematopoietic response in Rad21 +/2 mice.
We used a surrogate assay to measure stem-like cells by counting high proliferative potential-colony forming cells (HPP-CFC) in the hematopoietic stem cell compartment and the low proliferative potential-colony forming cells (LPP-CFC), characteristic of the committed progenitor cell pool [41]. Bone marrow cells from femurs of mutant and WT littermates were analysed at day 10 post 8Gy WBI. The result showed that LPP-CFCs were significantly reduced following IR, with approximately 0.1% cell survival. Although the average number of LPP-CFC in Rad21 +/2 mice appeared to be lower compared to WT, this was not significant so (p.0.05) (Fig. 7A). We noted that WT mice displayed greater individual variation in cloning efficacy than Rad21 +/2 mice. Similarly, the number of HPP-CFC in Rad21 +/2 mice was slightly but not significantly reduced compared to WT following IR (p = 0.205) (Fig. 7B). These data suggest impaired post-irradiation bone marrow stem cell clonogenic regeneration in Rad21 +/2 mice. It remains to be determined whether the bone marrow stem cells and progenitor cells recover to normal levels, or remain suppressed in Rad21 mutant mice. Successive transplantation models might be instructive in probing this possibility.

Identification of Rad21 cohesin as a novel mammalian radiation responsive gene
Pioneering studies established Ataxia telangiectasia patients as a most radiation sensitive group of patients [1]. Being an autosomal recessive disorder, individuals with two Atm mutant alleles manifest a broad spectrum of neurological, oncological and immunological pathologies but fortunately the population frequency of Atm patients is relatively low. However, Atm carriers are far more frequent (around 1:200) and it has been an ongoing concern that some of these individuals are likely to be IR sensitive even though this has been difficult to demonstrate [42]. In contrast, for a different endpoint, female Atm carriers have a clearly elevated risk for developing breast cancer [5][6][7]. Mouse models of Atm deficiency largely recapitulate the human AT phenotypes [28,43]. Hence, some pathologies in Atm patients and mouse models, are evident without overt IR exposure but unlike Rad21 homozygous mutant mice, they are viable. As Rad21 is an essential gene and Rad21 2/2 embryos die very early in development, we investigated mice with only one mutant allele. This proved very instructive. The observations made here with Rad21 +/2 mice indicate that one Rad21 allelic mutation confers a level of IR sensitivity that eclipses that observed in Atm heterozygous mice and accordingly raises the issue that should cancer patients with Rad21 or potentially other cohesin mutations be subjected to RT they may suffer adversely from standard RT protocols.
WBI studies initially highlighted a reduced animal survival in Rad21 +/2 mice. Furthermore, the absence of genetic interaction with Atm mutant mice is consistent with cohesin functioning downstream of Atm signalling [15,23,44]. More focused studies that explored tissues that impact tolerance doses of RT, such as the GIT and bone marrow, revealed that these tissues were indeed hypersensitive to IR. It was most apparent that Rad21 +/2 mice are hypersensitive to the toxic effects of RT when compared to wild type littermates and this was manifested by crypt losses in both SI and colon. Importantly, the kinetics of the responses were different in these two tissue compartments with the notable observation that in the colon of Rad21 +/2 mice the characteristic and early (within the first 24 hours) shut down of proliferation following IR was absent, which would allow cells to progress with the burden of potentially lethal DNA damage. Our in vivo characterisation of Rad21 +/2 mutants was thus reminiscent of a classical feature of Atm-mutant cells, that of checkpoint failure, including radioresistant DNA synthesis [45]. Collectively, these novel data indicate that Rad21 plays a fundamental role(s) in the IR response in whole mammals. Rad21 +/2 mutant animals are thus a unique resource for dissecting the molecular determinants of IR responses in the context of whole mammals.
Mouse Rad21 cohesin is required for HR Rad21 was implicated in promoting homologous recombination in fission yeast [8,13]. Although the role of yeast RAD21 cohesin in mediating chromosome segregation was found to be conserved in mammals, there has been no definitive data to support a role for mammalian RAD21 in HR. We found that a reduction of Rad21 gene dosage resulted in a significant reduction in the efficiency of gene targeting in mouse ES cells and in the frequency of MMCinduced SCEs in MEF cells, both processes mediated by HR [21,46]. Our finding provides the strong evidence that RAD21, and/or the cohesin complex, directly or indirectly promotes HR in mammalian cells. Consistent with this finding, we showed that Rad21 +/2 ES cells were exquisitely sensitive to MMC. The repair of MMC-induced DNA lesions involves multiple DNA repair pathways, with HR and other HR-dependent DNA repair pathways, i.e. Fanconi anaemia (FA) and nucleotide-excision repair (NER), being the main repair pathways [19]. The observed MMC sensitivity in Rad21 +/2 ES cells supports a role for RAD21 in HR. Furthermore, this result suggests that RAD21 may be involved in one or more these DNA repair pathways.
We did not observe a significant reduction in the survival of Rad21 +/2 ES and MEF cells compared to that of WT cells following IR treatment. However, we found that IR resulted in significantly higher level of cH2AX foci in Rad21 +/2 MEF cells when compared with WT cells, suggesting impaired DSB repair in Rad21 +/2 MEF. Further, Rad21 +/2 MEFs have an elevated level of NPB and MN following IR. NPB and MN have been shown to originate from abnormal chromosomal structures such as dicentric chromatids, ring chromatids and acentric fragments [47]. These abnormal chromosomal structures form when cells undergo DNA replication in the presence of unrepaired DSBs and repair after replication, resulting in misjoined broken chromosome ends [47]. It is therefore likely that higher levels of NPB and MN observed in Rad21 +/2 cells are a consequence of the impaired DSB repair in combination with failure in cell cycle arrest following IR. Consistent with this proposition, our results showed that a significantly higher percentage of Rad21 +/2 MEF cells continued to undergo DNA replication and mitosis following IR.
The absence of significant IR sensitivity in Rad21 +/2 cells is in contrast to the marked IR-sensitivity observed in Rad21 yeast mutants. The differences in DSB repair mechanisms between yeast and mammalian cells may explain, at least in part, the modest IR sensitivity in Rad21 +/2 cells. Unlike yeast which predominately uses a high-fidelity HR to repair DSBs [48], mammalian cells employ two distinct mechanisms for the resolution of DNA DSBs: the main pathway, non-homologous end-joining (NHEJ), and the other, the HR pathway [49]. HR is primarily used during S/G2 phase of cell cycle when sister chromatids are available as templates for repair. Although Rad21 +/2 cells are deficient in HR repair, it is possible that these cells use the NHEJ pathway to repair IR-induced DSBs, which in turn alleviates the IR sensitivity of Rad21 +/2 cells. Consistent with this proposition, a most recent study showed that radiation-induced DSB repair was impaired in G2, but not G1 phase in Rad21-depleted cells [50]. Further analysis using compound mutant mice models with either HR or NHEJ deficiency may assist in elucidating the role of cohesins in this context.

Failure of crypt regeneration underpins GI sensitivity in
Rad21 +/2 animals Radiation resulted in a higher percentage of the crypt death in the SI in Rad21 +/2 animals compared to WT. IR damage to the GIT in mice is thought to result from two main scenarios: (i) as a direct consequence of damage to epithelial crypt cells [51] or, a more recent and hotly debated suggestion (ii) as a result of damage secondary, for example, to endothelial cell apoptosis [52]. We consider the possibility of secondary damage unlikely, as no post-IR microscopic evidence of endothelial apoptosis was seen in mutant intestines. To more definitively examine which of these mechanisms may apply to Rad21 mutant cells, we performed an ex vivo crypt organoid assay; no angiogenesis occurs within these bodies. The data clearly showed that Rad21 +/2 crypt organoids are also more sensitive to IR than WT, without the possibility of endothelial-damage-generated consequences. Hence, the observed GI sensitivity in Rad21 +/2 mice is likely to be the result of direct damage to crypt epithelial cells.
Following radiation, a temporary mitotic arrest and subsequent cell death occurs [39]. These events comprise some of the key early responses of the GIT to radiation-induced damage. Cell death and lack of cell proliferation in combination with continued programmed cell shedding at the apices of villi results in extensive depletion of crypt cells, leading to crypt shrinkage and perturbed morphology [35]. Depending on the extent of damage, some shrinking crypts may resume proliferation within 24 hours to 36 hours post IR and those crypts may survive as a result of crypt regeneration, whereas others will ultimately die [31]. Our analyses revealed significantly less proliferative cells in the SI crypts of Rad21 +/2 mice 3.5 days post IR, suggesting that the process of crypt regeneration is severely compromised. We found that TUNEL-positive cells occurred more frequently at the crypt bases 6 hours post IR in Rad21 +/2 mice than that of WT animals. The base of SI crypts contains stem cells which are recruited from quiescence to a proliferative state in response to IR damage. Therefore, the lack of crypt regeneration in Rad21 +/2 mice could be due to the enhanced killing of their crypt stem cells by IR, leading to regeneration insufficiency of surviving clones. A second possibility exists: that unlike WT cells, the mutant crypt cells fail to withdraw from the cell cycle to repair lethal damage. This view was formed based on our results showing that the frequency of PCNA-positive cells at day 1 post IR in the irradiated Rad21 +/2 animals remained similar to that of unirradiated mice, suggesting ongoing active cell proliferation and failure to initiate cell cycle arrest. This proposition was further supported by our finding that Rad21 +/2 MEFs are defective in cell cycle arrest following IR. Cells with defective cell cycle checkpoints have been shown to be more susceptible to death by mitotic catastrophe, after treatment with DNA-damaging agents [53]. Mitotic catastrophe, also known as reproductive death, occurs because cells enter mitosis in the presence of residual or misrepaired DNA DSBs, leading to the propagation of mutations and chromosomal aberrations which ultimately results in cell death [54]. Further, we found that there was a higher incidence of cells with radiation-induced cH2AX foci at the crypt bases in Rad21 +/2 mice. We noted a higher level of apoptosis in Rad21 +/2 animals at later time-points (day 1 and day 3.5 post IR). It is therefore plausible that mitotic catastrophe in crypt cells is consequential to their DSB-induced proliferative response during crypt regeneration in Rad21 +/2 animals. Taken together, our data suggest that the GI sensitivity of Rad21 +/2 animals may be attributed to the impairment of crypt regeneration as a result of enhanced killing of crypt stem cells in combination with a failure in mitotic arrest followed by mitotic catastrophe.
It is intriguing that the severe sensitivity of Rad21 +/2 animals is not reflected in Rad21 +/2 ES and MEF cells. Multiple factors, such as the difference between cellular-and tissue-specific responses, may have contributed to the discrepancy. For example, small intestinal radiosensitivity is known to be largely associated with the rapid turn-over of small intestinal crypt cells [29]. Unlike ES and MEF cells, the epithelium (crypts and villi) of small intestines consists of cells that are short-lived and extruded within 3-5 days as they migrate up to the tip of villi. The renewal of crypt epithelial cells largely depends on a sub-population of stem cell progenitors residing at the base of crypts. Our results showed that following IR, Rad21 +/2 crypt cells fail to undergo cell-cycle arrest and continue to migrate towards the tip of crypts where they will be shed. Furthermore, we showed that the stem cell progenitors of Rad21 +/2 crypts are more susceptible to IR-induced cell death. The combination of these two factors inevitably results in the accelerated loss of crypt cells, leading to the enhanced crypt death. Presumably, it is these unique features of small intestinal crypts that revealed the severe intestinal radiosensitivity of Rad21 +/2 animals.
GI sensitivity of Rad21 +/2 mutant animals -relevance to human radiation responses The systematic and mechanistic investigation of Rad21's role in the radiosensitivity of different normal tissues will be important for understanding the radiation responses in both tumours and normal tissues. Malignant tumors typically have a high proliferative fraction and in some ways behave similarly to tissues with rapid cell turnover, such as the SI. Our finding that SI crypt cells and long-lived hematopoietic cells of Rad21 +/2 mutant animals are more susceptible to killing by radiation raises the possibility that targeted depletion of Rad21 in tumors would be of therapeutic utility and that the level of differential depletion between normal adjacent tissue and tumor cells would not need to be absolute to achieve a clinically relevant differential tumor cell kill. Conversely, our study has another important clinical implication. Cancer patients who sustain greater than expected side-effects in the SI from RT, may harbour mutations in, or dysregulation of, Rad21.
Clinical RS and its unfortunate functional consequences for cancer patients also provide a rationale for a predictive assay for testing for Rad21 status prior to RT, as a form of molecularly individualized medicine. Those with Rad21 defects could be spared irradiation or their doses limited, while WT cases could be considered for RT dose intensification. Given the typically steep dose-response curve for tumour cures by RT, the latter could conceivably result in enhanced cancer cures. Furthermore, human populations are expected to harbour a generally greater likelihood of heterozygous mutations than their quadratically-less-likely homozygous counterparts. Our finding of a gene dosage effect for mammalian Rad21 IR sensitivity thus suggests that if Rad21 function is conserved in mammals, Rad21 DNA sequence variants, if present, should be discernable from background.

Materials and Methods
Targeted deletion of the Rad21 gene and generation of mutant animals Gene targeting and generation of mutant mice were carried out essentially as described previously [25]. Briefly, to construct the targeting sequence, 59 and 39 targeting arms of 1.5 kb and 5.0 kb, respectively, were generated by PCR from the W9.5 ES cell line [55]. The targeting construct was made by fusing a splice acceptor (SA) with an IRES neo gene cassette which has the SV40 polyA sequence to terminate the mRNA. The resulting targeting construct replaced the exon 2 of the Rad21 gene, resulting in a null allele (Fig. S1A). Linearized DNA was introduced into W9.5 parental ES cells by electroporation and positive clones selected as described previously [25]. Protein expression in WT and three neomycin-resistant ES clones was examined by Western blot analysis. A ,120 kDa Rad21 protein band was detected in each cell line (Fig. S1B). Judging from signal intensities, the clone 5 displayed a noticeably reduction of RAD21 protein level, while two other clones (83 and 96) had similar signal intensity as WT. The level of PCNA was similar in all cell lines, indicating that any the reduction of the RAD21 protein is not necessarily associated with changes in expression of this protein that is typically associated with cell proliferation. Targeting of this clone was confirmed by PCR and Southern blot analysis of genomic DNA. DNA Sequence analysis of PCR products confirmed the insertion of the neo gene into the correct site of the Rad21 gene. Accordingly, ES clone 5 was used for microinjection into C57Bl/6J blastocysts. Chimeric offspring were obtained. Heterozygous Rad21 +/2 mice were generated by breeding. Tail DNA was used for PCR-based genotyping.
Cell culture, ES cell targeting efficiency and mouse embryonic fibroblast cell lines For the targeting efficiency assay, ES cells were cultured in BRLconditioned medium as described elsewhere [46]. WT and the Rad21 +/2 ES cell lines were transfected with a Rad54-GFP knockin construct [20]. One week after selection with Puromycin, single-cell suspensions of surviving colonies were made following trypsinization and analysis by FACS on a green fluorescence (eGFP) versus forward scatter (FSC-H) plot. Results were plotted in a fluorescence (GFP) histogram as described previously [20]. Targeting efficiencies were determined using Modfit software (Beckton Dickinson) and are indicated in each histogram. Rad54 2/2 ES cell lines were used as gene targeting-defective controls.
MEF cell lines were established from E13.5 embryos as described [55]. The procedure was approved by the Peter MacCallum Cancer Centre Animal Experimental Ethics Committee (approval #1275 and #1346). Cells are maintained in DMEM supplemented with 10% FCS, 1% Penicillin and Streptomycin, and 0.2% 0.1 M b-mercaptoethanol.

Clonogenic survival assays
WT and isogenic Rad21 +/2 ES cells were seeded in 35 mm plates at appropriate densities in duplicate plates following a single dose of graded c-ray irradiation (0, 2, 4, 6 and 8Gy). Cells were cultured for 14 days and colonies containing more than 50 cells were scored as clonogenic survivors. For assaying MMC sensitivity, ES cells were seeded in 6-well plates at appropriate densities, allowed to adhere and treated with MMC for 3 hours. Following treatment, MMC was removed and plates washed. Cells were cultured in MMC-free medium for 14 days and surviving colonies of more than 50 cells were counted. Three independent experiments were performed for each assay.

FACS analysis
MEFs were irradiated at 10Gy and harvested at 24 and 48 hours post IR. Unirradiated cells were used as controls. Cells were fixed in 70% ethanol and treated with propidium iodide and RNase A. DNA content was analysed by flow cytometry. For BrdU labelling, cells were incubated in 10 ìM BrdU for 45 min prior to harvesting. Cells were then fixed in 70% ethanol, washed and treated with 2 N HCl containing 0.5% TritonX-100 for 30 minutes. Following neutralisation with 0.1 M Na 2 B 4 O 7 , cells were incubated with a monoclonal anti-BrdU antibody (BD Biosciences) for 30 minutes. Alexa 488 anti-mouse IgG (Invitrogen) was used as the secondary antibody. Cells were stained with propidium iodide and analysed by flow cytometry. Cell cycle profiles were analysed using ModFit and FCS express 3 softwares.

Chromosome spreads, SCE assay and Immunofluorescence
Metaphase chromosome spreads were prepared as described previously [56]. SCE was carried out essentially as described elsewhere [57]. Briefly, MEFs were cultured in the presence of 5 mM BrdU for 48 hours. Cells were then incubated in Colcemide (0.15 mg/ml final concentration) for 2 hours and chromosome spreads were prepared. For MMC-induced SCEs, cells were treated with MMC (6 mM final concentration) for 30 minutes prior to the addition of Colcemide. SCEs were scored in a minimum of 20 metaphase spreads with near-diploid chromosome content. Immunofluorescence was performed as described previously [56]. For cH2AX immumostaining, MEFs were irradiated at 10Gy and fixed 4 hours post IR. Unirradiated MEFs were used as basal controls. Primary antibodies were a rabbit polyclonal anti-cH2AX (gift from Dr W. Bonner) and a mouse monoclonal anti-a-tubulin antibodies (Sigma-Aldrich, used at 1:1000 dilution). Secondary antibodies were Alexa-488 anti-mouse and Alexa-568 anti-rabbit antibodies (Invitrogen). Cells were counterstained with either DAPI or propidium iodide. Images were collected using an Olympus BX51 microscope and cH2AX foci were counted using Metaphorph software.

Whole body irradiation
Animals of mixed (129/Sv X C57BL/6) background were housed in micro-isolators in a specific pathogen free (SPF) facility. Compound mutant mice of Rad21 +/2 and AtmSRID [28] were generated by intercross. Adult mice (eight to ten week-old littermates) were exposed to a single dose of WBI (6Gy, 8Gy, 10Gy or 13Gy) using a Caesium-137 gamma source at the dose rate of 1.67 min/Gy. Following IR, mice were given water containing antibiotic (0.5 g/L Neomycin), except those used for bone marrow assays. Animals were checked daily for signs of stress, including abnormal posture (head down), ambulation (reluctance to move and unresponsive to activity), laboured breathing (panting or gasping), loss of appetite, rough coat and diarrhoea. Animals displaying signs of severe stress were culled. Animals used for tissue analysis sensitivities were sacrificed at specified time-points post WBI. For determining survival, animals were observed for 30 days. The procedure for animal experiments was approved by the Peter MaCallum Cancer Centre Animal Experimental Ethics Committee (approval #1275 and #1346).

Histology
Mice were sacrificed by cervical dislocation following required time-points following WBI. Twelve segments of 0.5 cm in length were collected from the SI jejunum, fixed in 10% normal buffered formalin, embedded in paraffin and sectioned transversely. Because of the variation in radiation response in different regions of the colon, only the distal colon was used for analysis and sections were cut longitudinally [31,51]. Sections were stained with hematoxylin & eosin (H&E).

Microcolony assay
SI samples were collected at day 3.5 post WBI following various doses and processed asdescribed above. For each mouse, the number of surviving crypts was scored in at least six whole transverse SI sections according to the criteria set for microcolony assay [30,31]. A surviving crypt was defined as containing 10 or more darkly stained nuclei, little cytoplasm and lying close together. Scoring was performed in a code-blinded fashion with samples identified only by histological block number. SI's from at least five mice were scored per data point.
Phospho-histone H3-positive cells were counted in the whole circumference of at least four SI cross-sections. MI was derived from the number of mitoses per crypt. PCNA-positive cells in large intestinal crypts were recorded on a positional basis using a central cell at the crypt base as position 1. Accordingly longitudinal crypt positions on both walls of the crypt can be assigned numbers 1, 2, 3 etc. and pooled for scoring purposes [38]. Only cells stained with strong intensity were scored. All other cells that stained either weakly or not at all were regarded as negative. At least 20 crypts per mouse were scored and a minimum of five mice per data point were assessed. Counting was done in a code-blinded fashion, with samples identified by the histology block numbers and genotypes only being revealed at the end of scoring.

TUNEL staining
TUNEL staining, which detects DNA fragmentation in cells undergoing apoptosis, was done using the Apop-Tag Kit (Chemicon) following the manufacturer's instructions. Sections were counterstained with hematoxylin. Total numbers of TU-NEL-positive cells in SI crypts were counted in the whole circumference of at least four cross-sections in a code-blinded manner. The apoptotic index describes the average number of apoptotic cells per crypt. The frequency of TUNEL-positive cells was determined by scoring at least 20 crypts per mouse and three animals per genotype.

Colony forming cell (CFC) assay
Littermates were treated with a single dose of 8Gy WBI and sacrificed at day 10 post IR, along with unirradiated littermates. Bone marrow cells were harvested from one femur. Cells were plated in triplicates at 2,500 cells per 35 mm dish for unirradiated mice, or 1/10 and 1/100 per femur for irradiated mice using a double layer agar containing colony stimulating factor-1 (CSF-1) at 1610 3 U/dish), IL-3 at 25 U/dish, Interleukin 1alpha (IL-1alpha) at 800U/dish and stem cell factor (SCF) at 100 ng/dish [58]. Partially purified mouse uterus extract (PMEU) was used as a source of CSF-1. IL-3 was obtained from conditioned medium from a mouse mammary cell line. Plates were incubated in a 37uC incubator under 5% O 2 , 10% CO 2 and 85% N 2 for 14 days. Colonies (.50 cells/colony) were counted under an inverted microscope.
LPP-CFCs form, in the presence of CSF-1 alone, colonies containing 50 to 50,000 cells with a diameter of less than 0.5 mm. HPP-CFCs form, in the presence of multiple growth factors (i.e. CSF-1, IL-3, IL-1alpha and SCF), colonies of more than 100,000 cells with a diameter of greater than 0.5 mm. Colony number per mouse femur was determined. Colonies larger than 0.5 mm diameter were scored as HPP colonies and those less than 0.5 mm in diameter as LPP colonies.

Western blot analysis
Protein extracts were prepared as described [56]. Western blots were probed with a polyclonal anti-Rad21 antibody (Abcam, used at 1:500 dilution) and a polyclonal anti-SMC3 antibody (Abcam, used at 1:1000 dilution). Signal was detected with chemiluminescence. The membrane was then stripped to remove any signal and re-probed with a monoclonal anti-PCNA antibody (DAKO) and a monoclonal anti-a-tubulin antibody (Sigma-Aldrich).

Statistical analysis
Statistical analysis was performed using GraphPad Prism 5 software. Clonogenic survival post IR was analysed using the linear quadratic model and survival post MMC treatment was analysed using a two components exponential model. P values were determined using the Students T-test, one-way ANOVA test or Chi-square test. The Log-Rank test was used for cumulative survival analysis.