|Home | About | Journals | Submit | Contact Us | Français|
Recent studies suggest that PARP1 inhibitors, several of which are currently in clinical trial, may selectively kill BRCA1/2 mutant cancers cells. It is thought that the success of this therapy is based on immitigable lethal DNA damage in the cancer cells resultant from the concurrent loss or inhibition of two DNA damage repair pathways: single-strand break (SSB) repair and homologous recombination repair (HRR). Presumably, inhibition of PARP1 activity obstructs the repair of SSBs and during DNA replication, these lesions cause replication fork collapse and are transformed into substrates for HRR. In fact, several previous studies have indicated a hyper-recombinogenic phenotype in the absence of active PARP1 in vitro or in response to DNA damaging agents. In this study, we demonstrate an increased frequency of spontaneous HRR in vivo in the absence of PARP1 using the pun assay. Furthermore, we found that the HRR events that occur in Parp1 nullizygous mice are associated with a significant increase in large, clonal events, as opposed to the usually more frequent single cell events, suggesting an effect in replicating cells. In conclusion, our data demonstrates that PARP1 inhibits spontaneous HRR events, and supports the model of DNA replication transformation of SSBs into HRR substrates.
Poly (ADP-ribosyl)ation is the posttranslational transfer of long chains of negatively charged ADP-ribose moieties to proteins. The resultant increase in negative charge causes the target protein to lose DNA-binding affinity (1). Poly (ADP-ribose) polymerases, or PARPs, comprise a large family of genes that have shared homology with the catalytic domain of the founding member, PARP1 (1). PARP1 has been widely implicated in a multitude of cellular processes including replication (2–4), transcription [reviewed in (5)], chromatin remodeling [reviewed in (5)], telomere maintenance (6) and perhaps most notably, the repair of DNA damage through the base excision repair (BER) pathway (7–9).
Current understanding is that the key BER proteins actually participate in several distinct pathways such as short-patch BER, long-patch BER, single strand break (SSB) repair and nucleotide incision repair (10). However, the common factor for all of these pathways is an SSB—be it the initiating lesion or an intermediate step in a repair process. PARP1 readily binds SSBs (11,12) and recruits the scaffolding protein XRCC1 (13). PARP1 poly (ADP-ribosyl)ates itself (13), reducing its DNA-binding affinity, thus allowing other repair factors to bind the lesion site (9,14).
A recent study demonstrated that chemical inhibition of PARP1 decreased the efficiency of SSB repair (15), conjecturing that chemically inhibited PARP1 remains bound to DNA and blocks other repair proteins from the SSB site. However, the same study revealed that despite PARP1 silencing via RNA interference, cells were able to repair SSBs (15), indicating that an alternative pathway, possibly homologous recombination repair (HRR), can compensate for this loss. Loss of Parp1 by way of gene targeting in human cells does not hinder formation of nuclear RAD51 foci (an indicator of RAD51-dependent HRR) (16), nor does PARP1 inhibition appear to obstruct HRR in vitro (16,17). Waldman and Waldman (18) found a 4-fold increase in intrachromosomal homologous recombination in mouse fibroblasts grown in the poly(ADP-ribose) polymerase inhibitor, 3-methoxybenzamide, compared to controls. Furthermore, PARP1 does not co-localize to RAD51 foci following DNA damage (16) indicating that it is unlikely that PARP1 is directly involved in the HRR process. In addition, increased sister chromatid exchange has been observed with PARP1 inhibitors in Chinese hamster ovary cells (19) and in PARP1 null mice (20), whereas over-expression of Parp1 decreases the incidence of sister chromatid exchange following DNA damage (21). Resolution of SSBs by way of HRR in the absence of PARP1 activity may be due either to stalled replication fork or DSBs resulting from replication fork collapse. The requirement of such activity is the postulated basis for synthetic lethality observed when treating breast and ovarian cancer cells deficient for either BRCA1 or BRCA2 with PARP1 inhibitors (4,22–24). This is because BRCA1 and BRCA2, amongst their various functions, are required for RAD51 dependent double-stranded DNA break induced HRR (25–27). Together, these in vitro observations indicate that loss or inhibition of PARP1 leads to a hyper-recombinogenic phenotype.
Here, we evaluate the spontaneous frequency of HRR in vivo using the well-established and highly sensitive pun eye spot assay (28–30). The murine pink-eyed dilution gene, p, encodes a protein that functions in the pigmentation of the fur and the retinal pigment epithelium (RPE) of the mouse (31). Mice lacking a functional copy of this gene are hypopigmented, having a dilute (gray) coat and pink eyes (the cells of the RPE are rendered transparent) (31). One such mutant, the ‘pink-eyed unstable’ (pun) allele contains a 70-kb duplication of exons 6–18 (32–34, Figure 1) and causes this autosomal recessive phenotype. However, the pun allele is subject to a relatively high frequency of spontaneous, somatic reversion to wild-type (35). Reversion can only be attributed to HRR mediated deletion of the duplicated exons, which restores functionality of p (32,33) and consequently pigmentation to the fur and RPE. Equivalent assays in yeast have demonstrated that such intrachromosomal deletions between homologous tandem repeats may be mediated by either a RAD51-dependent pathway (canonical HRR pathway) or a RAD51-independent pathway [single strand annealing (SSA), an alternative HRR pathway)] (36). Therefore, the frequency of pun reversion is indicative of the somatic occurrence of spontaneous HRR events (28,29,37). Here we use the pun eye spot assay to demonstrate that the absence of PARP1 results in increased spontaneous somatic HRR events in vivo. The significant increase in rare multi-cell clones of eye spots in Parp1 nullizygous mice suggests that the normal function of PARP1 is to remove DNA lesions prior to their becoming HRR substrates during DNA replication. Our observations substantiate current models of the relationship between PARP1 and HRR, providing formal in vivo evidence of a spontaneous, hyper-recombination phenotype.
Mice heterozygous for a targeted null allele of Parp1, 129S-Parp1tm1Zqw (38), were obtained from Jackson Laboratories (Bar Harbor, ME) and genotyped for Parp1 as described (http://parplink.u-strasbg.fr/protocols/tools/parp1_typing.html) earlier. The Parp1+/− mice were made C57BL/6J congenic by five backcrosses followed by two crosses to C57BL/6J pun/un mice. Mice with homozygous pun/un genotype were selected by their phenotypic gray coat color. The resulting C57BL/6J N7 Parp1+/− pun/un mice (hereafter referred to as Parp1+/−) were self-crossed to create the necessary experimental (Parp1−/−) and control animals (Parp1+/− and Parp1+/+).
The pun frequency assay was performed as described earlier (29). Briefly, eyes were harvested, with the investigator blinded to the genotypes until after pun eye spot data was collected. Three types of data were collected for each RPE: the total number of eye spots, the number of cells comprising each eye spot, and the position of each eye spot relative to the optic nerve. Following Bishop et al. (29), a pun reversion event or eye spot was defined as ‘a pigmented cell or a cluster of pigmented cells, separated by no more than one unpigmented cell’. Eyes were viewed at 15× using a Zeiss Axiovert microscope and methodically scanned for pigmented spots. A 5× mosaic photograph was taken of each RPE using a Zeiss Lumar V.12 stereomicroscope, Zeiss Axiovision MRm camera and Zeiss Axiovision 4.6 software (Thornwood, NY, USA). In cases where a suspected clone consisted of pigmented cells separated by clear cell(s), the Adobe Photoshop CS2 (San Jose, CA, USA) measure tool was used to assess if the unpigmented area between pigmented cells was consistent with the single cell diameter in that part of the RPE (Figure 1). The pun positional assay was then performed as described earlier (28). Briefly, the position of each spot relative to the optic nerve was calculated using simple measurements. In Photoshop, the brush tool was used to mark the approximate center of the optic nerve and the measure tool was then used to obtain two distances for each eye spot: (i) from the center of the optic nerve to the proximal edge of the eye spot, and (ii) from the center of the optic nerve to the edge of the RPE. By dividing the former by the latter, the position of each eye spot relative to the optic nerve was determined.
Primary mouse embryonic fibroblasts (MEFs) were obtained by intercrossing Parp1 heterozygous mice to obtain Parp1 null embryos and littermate controls. Pregnancies were timed by the observation of a vaginal copulation plug and embryos were harvested on Day E14. Embryos were mechanically homogenized and allowed to incubate in 0.05% Trypsin–EDTA for 20min. Cells were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10000U/ml penicillin, 10000µg/ml streptomycin, 25µg/ml Amphotericin B (Cellgro). The cells were grown at 37°C in the presence of 5% CO2 in a humidified incubator.
HRR frequency was determined by immunofluorescence using an antibody against RAD51 (RK-70-005, MBL) as described earlier (39). Cells were grown on acid-washed fibronectin coated cover slips at a density of 1×105 cells/cover slip. Cells were then fixed with 4% paraformaldehyde and permeabilized using Triton X-100 for 10min at room temperature. Blocking was performed with 1% BSA–4% goat serum for 1h followed by an overnight incubation with RAD51 antibody (1:2000) in a humidified chamber. Goat antichicken Alexa Fluor 488 (Invitrogen) diluted 1:1000 was applied to each cover slip for 1h followed by DAPI staining and Fluoromount G slide mounting. A minimum of 100 nuclei were examined for each genotype, repeated with three technical replicates per genotype.
Parametric analysis of variance was performed in Microsoft Excel 1994 for Mac (Redmond, Washington). The Fmax test was done by hand per Hartley (40) and Rohlf and Sokal (41). Measurements of skew and kurtosis were obtained using descriptive statistics in Microsoft Excel 1994 for Mac (Redmond, Washington). Non-parametric Kruskal–Wallis test was performed using Stata (College Station, TX, USA). Dunn’s test was done by hand per Siegel (42). Chi-squared contingency analysis was performed using the VasserStats online calculator (http://faculty.vasser.edu/lowry/VasserStats.html, accessed 11/2009). Fisher’s Exact test (43) was used to compare RAD51 foci quantification using the sum of the technical replicates per genotype.
PARP1, amongst its various activities, is involved in SSB repair (11,12). Inhibition of the PARP1 protein in BRCA1 and BRCA2 mutant cells leads to their selective cell death (23,44). This has led to the working model that the observed cell death is due to immitigable DNA damage caused by the loss of two key DNA repair pathways: (i) HRR, due to loss of BRCA1/2, and (ii) SSB repair due to inhibition of PARP1 (4,23,24). This model suggests that lack of PARP1 alone will cause an increase in HRR frequency in the absence of exogenous damage. To test this assumption, we determined the spontaneous frequency of HRR in Parp1−/−, +/− and +/+ animals using the in vivo pun assay.
The frequency of pun reversion was determined in Parp1−/−, +/− and +/+ animals using pun eye spot assay (Table 1 and Figure 2). There is a clear increase in the frequency of pun reversion events observed in the absence of PARP1 compared to wild-type and heterozygous littermate controls. Considering that the data is non-parametric and not normal, we used the non-parametric Kruskal–Wallis test (followed by Dunn’s test to determine which groups are different from one another). This analysis revealed that Parp1−/− animals had a highly statistically significant increase in pun reversion, and thus increased HRR, compared to controls (P=0.0001).
The RPE develops radially outward from the optic nerve during development (45), therefore, much like the concentric annual rings of a tree, the positions of eye spots indicate the developmental time at which they occurred (28,29). pun reversion events closer to the optic nerve occurred earlier in development and events further from the optic nerve occurred later in development (29). To investigate whether the increased number of events occurred at a particular time during development, we analyzed the relative positions of the pun reversion events on the RPE of differing Parp1 genotype (35 wild-type, 28 heterozygous and 24 Parp1 null RPE). A Kruskal–Wallis test was used to compare the positional distribution of spots and indicated that there is a significant difference between genotypes (P=0.0043, Figure 3). Subsequently, a Dunn’s test indicated that the Parp1 null group is different from the control groups.
To determine whether the difference in the positions of spots was due to early or late events, comparison was done between the numbers of events with proximal positions (earlier in development, 0–0.50) versus the numbers of events with distal positions (later in development, 0.51–1.0). A 2×3 contingency table analysis comparing wild-type, heterozygous and null revealed a highly statistically significant difference in the proportion of spots on the proximal half of the RPE (χ2=32.09, P<0.0001). To verify that the difference was due to the null group, groups were analyzed pairwise. There was no significant difference between wild-type and heterozygous (P=0.865), whereas the null group was significantly different from each of them (P=0.005 and P=0.004, respectively). Finally, 2×2 analysis was performed comparing the null group against the combined heterozygotes and wild-type, confirming a highly statistically significant difference (χ2=17.86, P<0.0001), indicating that there is a shift of pun reversion events to earlier in development. One interpretation of this result is that there is an increased rate of homologous recombination that occurs in the absence of PARP1.
On initial analysis of the RPE, there appeared be a greater number of large (consisting of multiple cells) spots in the Parp1−/− RPE (Table 1). Previous studies have shown that spots with greater than 10 cells are very rare (28,30, Bishop, unpublished data). Therefore, a 2×3 contingency table analysis (Chi-square) was computed comparing the number of spots with 10 or fewer cells and the number of spots with 11 or more cells between genotypes (χ2=32.65, P<0.0001, Figure 4). To determine if the null group is the cause of statistical significance and because our frequency and positional analysis showed no difference between wild-type and heterozygous, the data for these two groups were combined and compared against the null group in a 2×2 contingency table analysis (χ2=28.64, P<0.0001). These results indicate that there is a significant increase in the incidence of large spots in the Parp1 null mice compared to controls, indicating an increase in HRR in proliferating cells.
To investigate whether a particular subset of spots (single cell or multi-cell) caused the observed shift in spots to an earlier time in development in Parp1 null mice, a Kruskal–Wallis test was used to analyze the relative positions of these subsets of eye spots between genotypes. This test revealed that multi-cell eye spots are shifted toward the optic nerve in the null mice as compared to heterozygous and wild-type (P=0.0001). Furthermore, analysis of the single-cell eye spots showed no difference between groups (P=0.5762, Kruskal–Wallis test), indicating that the positional shift observed when analyzing all spots is in fact due to the multi-cell eye spots. Considering that the Parp1 null mice display both an increase in the large eye spots and the position of these eye spots, we examined whether these large eye spots are significantly increased in the proximal half of the RPE in the absence of PARP1. A 2×2 contingency table analysis revealed that indeed there was a significant increase in large, multi-cell events (≥11 cells) in the proximal half of the RPE in null mice compared to control RPE (χ2=5.49, P=0.019, Figure 3). This suggests that many more highly replicative cells are prone to pun reversion events during early RPE development in the absence of PARP1.
It has been previously reported that human cells either deficient or inhibited for PARP1 have increased RAD51 nuclear foci, an indicator of increased HRR (16,17). To demonstrate that the same cellular phenomenon is observed here with the mouse Parp1 knockout model, we examined spontaneous levels of RAD51 nuclear foci in MEFs isolated from these mice (Figure 5). As expected, the observed frequency of spontaneous RAD51 foci for wild-type MEFs displayed a very low background, mostly in the category of 0–5 RAD51 foci per cell. However, Parp1 null MEFs showed a high number of cells with RAD51 foci with 68.3% having >6 foci per cell compared to wild-type MEFs (23.5%) (P=2.6e–29). In addition, comparing 0 foci to either the 6–10 or >10 foci per cell groups demonstrates a significant increase in numbers of foci per cell in the Parp1 null cells (P=3.3e–15 and P=8.3e–25, respectively). This data suggests a significantly higher frequency of HRR in Parp1 null MEFs compared to wild-type, correlating well with our in vivo pun assay observations.
The inhibition of PARP1 activity is an exciting novel therapy used in the treatment of BRCA1 and BRCA2 hereditary breast and ovarian cancers (22). The mechanism by which this therapy is thought to work is by synthetic lethality resulting from an inability to repair SSBs by either PARP1-dependent SSB repair or BRCA1/2-dependent HRR (after replication fork stall/ collapse) (4,23,24). This model would suggest an increase in the spontaneous frequency of HRR in the absence of PARP1. In vitro evidence for a hyper-recombination phenotype has been presented in various tissue culture experiments (16,17,19–21). Here, we provide in vivo evidence that spontaneous HRR frequency is indeed increased in the absence of PARP1 protein by using an established mouse model for measuring HRR events. HRR dependent repair of SSBs has been proposed to be due to DNA replication fork collapse at the SSB that converts these lesions into HRR substrates. Our data supports this notion, with a significant increase in large clonal eye spots in the Parp1 null background compared to controls. There is no selective advantage for pun reversion, thus it is likely that similar somatic homologous recombination events are occurring in replicating cells throughout all tissues of the body and not just in the developing RPE and at the pun locus.
The pun HRR assay is based on the loss of one copy of a DNA duplication that encompasses exons 6–18 of the p gene, restoring the function of this pigmentation gene. Theoretically the pun reversion event, a homology directed deletion, may be mediated by either of two alternative mechanisms (36). First, there is the canonical, RAD51/BRCA1-dependent HRR pathway and is likely instigated by either a double-stranded DNA break or DNA replication fork collapse. The alternative homology-dependent repair pathway is SSA. SSA is a RAD51-independent pathway that can mediate intrachromosomal deletions between homologous DNA sequences, but is unlikely to be dependent upon the DNA replication process and is more likely to act in response to a double-stranded DNA break than an SSB-induced replication fork collapse (46,47). In contrast, RAD51-dependent HRR is considered a high fidelity DNA repair pathway and is likely a favored mechanism in replicating cells. Therefore, RAD51-dependent HRR provides the most obvious explanation for the observation that earlier pun reversion events in proliferating cells will lead to larger eye spots.
Using the pun assay, we previously reported that there is a significant increase in HRR in Atm, p53 and Gadd45a null mice compared to controls (28). While these models showed differing shifts in the timing of events (position of eye spots) (28), the distribution of eye spot size was equivalent to wild-type (a majority of single cell eye spots, fewer two cell eye spots, and so forth, Bishop, unpublished data). In contrast, in the Parp1 null mice we observed a significant increase in the number of multi-cell spots, particularly those greater than 10 cells (Figure 4). It is possible that different HRR pathways are at work in these different mutant mice, with a preference for replication-tied RAD51-dependent HRR events observed in PARP1 null mice.
It has already been postulated that the mechanism by which Parp1 nullizygosity increases HRR is through replication fork collapse. Therefore, it is probable that it is the RAD51/BRCA1-dependent HRR pathway that results in the hyper-recombination phenotype we have observed in the Parp1 null animals. In such a case, it would logically follow that more of the spots consist of a greater number of cells because the original cell in which the pun reversion occurred was a proliferating cell; the daughter cells would have inherited the reverted p allele and thus will also be pigmented. However, in wild-type, as well as in Atm, p53 and Gadd45a null mice, the majority of spots are single-cell events. These events follow the same general pattern of positions as larger spots, but with a phase shift toward the optic nerve. This relative distribution suggests that the single-cell events were likely to have occurred in cells that were in their terminal division at the rear of the proliferating region of the RPE (28). Therefore, it is possible that these single-cell events were not necessarily tied to active replication machinery, but rather could be due to SSA, in a DNA replication-independent manner. BRCA1-dependent HRR is thought necessarily to involve RAD51 (48) yet SSA is thought to be a RAD51-independent event (36). In support of this, we observe a significant increase in spontaneous nuclear RAD51 foci in the absence of PARP1. Though we cannot determine the proportion of pun reversions that result from SSA events, it appears that the absence of PARP1 results in a clear increase in pun reversion events that are tied to cellular proliferation and DNA replication. Overall, our observations provide formal evidence that the absence of PARP1 protein results in a spontaneous hyper-homologous recombination phenotype that supports the proposed mechanisms of PARP1 inhibition and BRCA1/2 null synthetic lethality.
National Institutes of Health (K22ES012264 to A.J.R.B.); US Department of Defense Breast Cancer Research Program Predoctoral Traineeship Award (W81XWH-10-1-0026 to B.K.); Greehey Children’s Cancer Research Institute’s summer program (to C.B.). Funding for open access charge: Greehey Children’s Cancer Research Institute.
Conflict of interest statement. None declared.
The authors thank Ting Ting Gu for general technical support and Jo Ann Martinez for helping to establish the mouse lines. The authors also thank members of the Bishop lab for critical reading of the manuscript.