PARP-1 cooperates with Ptc1 to suppress medulloblastoma and basal cell carcinoma

The patched (Ptc1) protein is a negative regulator of sonic hedge- hog signaling, a genetic pathway whose perturbation causes de-velopmental defects and predisposition to speciﬁc malignant tumors. Humans and mice with mutated Ptc1 are prone to medulloblastoma and basal cell carcinoma (BCC), both tumors showing dependence on radiation damage for rapid onset and high penetrance. Poly(ADP-ribose) polymerase (PARP-1) is a nu- clear enzyme that plays a multifunctional role in DNA damage signaling and repair. In healthy and fertile PARP-1-null mice, radiation exposure reveals an extreme sensitivity and a high genomic instability. To test for interactions between PARP-1 and sonic hedgehog signaling, PARP-1-null mice were crossed to Ptc1 heterozygous mice. PARP-1 deletion further accelerated me- dulloblastoma development in irradiated Ptc1 1 / 2 mice, showing that PARP-1 inactivation sensitizes cerebellar cells to radiation tumorigenic effects. In addition to increased formation and slowed down kinetics of disappearance of g -H2AX foci, we observed increased apoptosis in PARP-1-deﬁcient granule cell progenitors after irradiation. Double-mutant mice were also strikingly more susceptible to BCC, with > 50% of animals developing multiple, large, inﬁltrative tumors within 30 weeks of age. The results provide genetic evidence that PARP-1 function suppresses sonic hedgehog pathway-associated tumors arising in response to environmental stress. To minimize mortality for medulloblastoma, two additional groups of PARP-1 þ / þ / Ptc1 þ / (cid:2) and PARP-1 (cid:2) / (cid:2) / Ptc1 þ / (cid:2) mice were irradiated at P1 with 3 mm thick lead shields positioned to protect mouse head during irradiation. Experimental protocols were reviewed by the Institutional Animal Care and Use Committee. and hybridized on normal metaphases obtainedfrom C57Bl/6 mouse embryonic ﬁbroblasts. Biotin and digoxigenin probes were revealed by Avidin–ﬂuorescein isothiocyanate (Vector Laborato-ries, Burlingame, CA) and anti-digoxigenin–rhodamine antibodies (Roche Diagnostics Corp., Indianapolis, IN), respectively. After DAPI staining, three-color images, from at least 15 well-hybridized metaphases, were captured using IPLab Spectrum software (BD Biosciences, Rockville, MD) from a cooled CCD camera mounted on a Microphot-FXA microscope (Nikon France S.A.S., Champigny sur Marne, FR). Classiﬁcation of G-banded chromosomes and the green-to-red ﬂuorescence ratios along individual chromo- somes were calculated by QUIPS-CGH software (Vysis, Downers Grove, IL). onset and development a Consequent increased incidence and more rapid onset of preneoplastic skin lesions in Ptc1 þ (cid:2) These show that PARP-1 to prevent skin cancer formation in response to DNA damage. PARP-1/ Ptc1 double-mutant here a unique new for of the by physical the study of the pathogenesis and treatment of this very common human

A number of PARP-1 knockout mouse models were generated by different groups (11)(12)(13). Despite its important role in cellular response to genotoxic stress, PARP-1 is not required for viability, and mice lacking functional PARP-1 develop normally and are not predisposed to early-onset tumors. PARP-1-null mice, however, show hypersensitivity to ionizing radiation and alkylating agents, and PARP-1-null cells exhibit chromosomal instability, shown by increased frequency of spontaneous sister chromatid exchange and DNA damage-induced micronucleus formation (12,14,15).
Growing evidence shows that PARP-1 activation plays an important role in the pathogenesis of several human diseases, such as stroke, myocardial infarction, circulatory shock, diabetes and neurodegenerative disorders, including Parkinson and Alzheimer (16)(17)(18). There is also evidence to implicate the role of PARP-1 activation in inflammatory disorders, such as arthritis, allergy, colitis and others (19). Thus, PARP-1 has been gaining increasing interest as a therapeutic target as PARP-1 inhibitors have been shown to be greatly effective in experimental models for some of these conditions. Several classes of PARP-1 inhibitors have been developed, and some of these are moving toward clinical development, or have already entered clinical trials for treatment of acute cardiac ischemia, or as chemotherapysensitizing agents (20).
On the other hand, several studies indicating important protective functions of PARP-1 in genome surveillance and DNA repair raise concern about the clinical use of PARP-1 inhibitors (21). Important issues are the potential side effects of long-term treatment and whether PARP-1 inhibition may increase the risk of mutagenesis or oncogenesis. While the therapeutic effects of PARP-1 inhibitors may exceed their potential risk for acute life-threatening diseases, development of PARP-1 inhibitors for inflammatory or neurodegenerative conditions may be more challenging because of the unknown potential long-term side effects of PARP inhibition.
Patched (Ptc1) heterozygous knockout mice (22,23), the mouse model for Gorlin syndrome, are cancer prone and hypersensitive to DNA damage caused by ionizing radiation that accelerates development of brain and skin malignancies, i.e. medulloblastoma and basal cell carcinoma (BCC) (24)(25)(26)(27). Development of both tumor types occurs through well-defined preneoplastic stages whose switch to malignancy could be driven by loss of the normal remaining Ptc1 allele (28,29). This suggests that processing of radiation-induced DNA damage is crucial for tumorigenesis in the Ptc1 þ/À model.
To investigate whether PARP-1 is an important genetic factor in tumor susceptibility, mice with a deletion in exon 2 of the PARP-1 gene (11) were intercrossed to Ptc1 þ/À mice, and the progeny was exposed to the damaging effects of radiation. Control and irradiated groups of Ptc1 þ/À mice with two, one or no intact PARP-1 alleles were placed on a lifetime study for development of brain and skin tumors or other neoplasms.

Radiosensitivity analysis
Brains (three/time point) of PARP-1 þ/þ /Ptc1 þ/À and PARP-1 À/À /Ptc1 þ/À mice irradiated with 3 Gy at P1 were collected 3 and 6 h post-irradiation and fixed in 4% buffered formalin. Serial sections of cerebellar tissues were cut at 4 lm thickness and stained with hematoxylin and eosin. Digital images from midsagittal cerebellar sections were collected by IAS image-processing software (Delta Sistemi, Rome, Italy). Cells showing signs of nuclear chromatin condensation and morphologically normal cells in the external granule layer (EGL) were counted using a double-blind method. Cell death was calculated as the percentage of pyknotic nuclei relative to the total number of cells. The total number of cells examined ranged from 1.5 Â 10 4 to 2.4 Â 10 4 per time point.

Immunohistochemistry analysis
Immunohistochemical analysis of rabbit polyclonal antibody against cleaved caspase-3 (Cell Signaling Technology, Danvers, MA) on brain samples was carried out on 4 lm thick paraffin sections as described (24). Immunohistochemical analysis of monoclonal antibody against c-H2AX (Upstate Biotechnology, Lake Placid, NY; 1:200 dilution) was performed using the HistoMouse MAX Kit (Zymed Laboratories, San Francisco, CA) according to the manufacturer's instructions.

Detection and counting of c-H2AX
To study kinetics of c-H2AX formation and elimination in X-ray-irradiated granule cell progenitors (GCPs) from PARP-1-proficient and -deficient mice in situ, we counted fractions of c-H2AX-positive nuclei in the entire EGL from sagittal sections of P1 cerebellum (three/time point) at 0.5, 3 and 6 h after irradiation (3 Gy). Digital images from midsagittal cerebellar sections, covering the entire EGL, were collected by IAS image-processing software (Delta Sistemi). Cells showing c-H2AX signals and negative cells in the EGL were counted using a double-blind method. The number of cells with c-H2AX foci was calculated as the percentage positive cells relative to the total number of cells. The total number of cells examined ranged from 6.6 Â 10 3 to 1.5 Â 10 4 per time point.

Western blot analysis
Proteins from cerebella were normalized for concentration (Bradford assay, Bio-Rad Laboratories, Hercules, CA) and separated on sodium dodecyl sulfate-polyacrylamide gel electrophoresis. Immunoblot analyses were performed as described (27). Anti-b-actin antibody was used to control protein loading. Antibodies included rabbit polyclonal antibody against p53 (Novocastra Laboratories, Newcastle, UK) and phospho-p53 (Cell Signaling Technology) and monoclonal antibody against b-actin (Sigma-Aldrich, St Louis, MO), 1:1000 dilution.
M. Tanori et al. result in dramatic susceptibility to environmental exposures. We found a modest reduction in size of the cerebellum and disorganization of the internal granule layer in Ptc1 þ/À mice with intact PARP-1 ( Figure 1E and F). Structural alterations were detected in 33.3% (4/12) of PARP-1 þ/þ /Ptc1 þ/þ and in 28.6% (2/7) of PARP-1 þ/þ / Ptc1 þ/À mice ( Figure 1I). In contrast, cerebellar development was severely impaired in irradiated PARP-1 À/À mice ( Figure 1G and H). Regardless of Ptc1 genotype, we observed a drastic enhancement in incidence and severity of cerebellum disorganization and major cerebellar abnormalities in 75% (6/8) of PARP-1 À/À mice ( Figure 1I). Alterations included dramatic decrease in thickness of the internal granule layer and disorganization of Purkinje cells, which instead of their normal monolayer arrangement were scattered within the cortex ( Figure 1H). Thus, PARP-1 deficiency strongly influenced the degree of radiation-induced damage in the developing cerebellum and impaired cell repopulation and normal growth.

Early preneoplastic lesions in cerebellum
During post-natal cerebellar development, differentiating GCPs complete their migration from the external to the internal granule layer by 3 weeks of age. Persistence of ectopic EGL areas after this time is indicative of impaired migration or differentiation ability of GCPs, suggesting a preneoplastic condition (31). Young asymptomatic Ptc1 þ/À mice (3-7 weeks) develop a cerebellum phenotype characterized by ectopic EGL areas abnormally expanding as nodular formations on the surface of cerebellar lobules ( Figure 2A) (27,28). In this study, histological examination of 5-week-old cerebella showed the presence of preneoplastic areas in 12.5% (1/8) of Ptc1 þ/À mice ( Figure 2B). A major effect of PARP-1 genetic inactivation was the drastic increase to 71.4% (5/7) of Ptc1-associated early neoplastic lesions in PARP-1 À/À mice, a nearly 6-fold increase (P , 0.05). Therefore, lack of PARP-1 strongly contributes to the growth of hyperplastic areas in the cerebellum of young Ptc1 þ/À mice. Similar to previous results with Ptc1 þ/À mice (28), the number of microscopic proliferations was greatly increased in the cerebellum of Ptc1 þ/À mice with intact PARP-1 function (85.7%; 6/7) after DNA damage caused by irradiation. Instead, irradiation of PARP-1 À/À /Ptc1 þ/À mice did not induce further increase in number of early lesions (75%; 6/8), indicating that genetic damage leading to GCP hyperproliferation can be acquired by early PARP-1 À/À neural progenitors in the absence of exogenous DNA damage, with no further acceleration of the process by radiation ( Figure 2B). Consistently, no differences in expression of the cell proliferation marker Ki-67 were detected in lesions arising in unirradiated and irradiated PARP-1 À/À /Ptc1 þ/À cerebella ( Figure 2C and D).
Survival and Ptc1-associated tumorigenesis in crosses between PARP-1-and Ptc1-mutant mice The observation that PARP-1 plays a crucial role in protection from the damaging effects of radiation in cerebellum tissue, and in suppression of the initial steps of medulloblastoma growth in Ptc1 þ/À mice, prompted us to examine whether PARP-1 germ line inactivation might also influence long-term survival and tumorigenesis.
-proficient mice were highly statistically significant (P , 0.001) at all time points examined. Since kinetics of disappearance of c-H2AX closely parallels the rate of DSB repair (36,37), longer retention of c-H2AX foci in PARP-1-deficient cerebella indicates impaired repair ability of radiation damage.
Increased DNA damage response and apoptosis in PARP-1-deficient GCPs To further define the requirement of PARP-1 activity in early signaling of DNA damage response, we examined p53, which plays a critical role in the response of CNS to genotoxic insult. Phosphorylation of murine p53 at Ser18 promotes both the accumulation and functional activation of p53 in response to damage by radiation, triggering p53-dependent apoptosis and cell cycle arrest (38). We used western blot analysis to determine total and Ser18-p53 protein levels in the irradiated cerebellum of PARP-1/Ptc1 double mutants at 3 and 6 h post-irradiation ( Figure 4D and E). Densitometric immunoblot analysis at 3 h showed 1.8-fold increased phosphorylation of Ser18-p53 in the cerebellum of PARP-1 À/À /Ptc1 þ/À compared with PARP-1 þ/þ / Ptc1 þ/À mice. This difference decreased out to 1.1-fold at 6 h. Our results suggest that PARP-1 inactivation alters signaling repair after DNA damage in GCPs.

Discussion
In this report, we have introduced PARP-1 mutations onto a Ptc1 þ/À background to investigate potential genetic interactions between the DNA strand break-detecting PARP-1 enzyme and Ptc1 during development and tumorigenesis. We show that cooperation of DNA end processing and Ptc1 function is required to suppress tumors arising from perturbations of sonic hedgehog signaling. This was reflected as increased frequency of both early and fully malignant tumors, reduced tumor latency and as the occurrence of frequent multiple tumors in PARP-1/Ptc1 mutants compared with Ptc1 þ/À animals.
Ptc1 has an early role in CNS tumorigenesis as young Ptc1 þ/À mice show abnormal hyperplastic EGL regions, suggestive of a preneoplastic condition (29). PARP-1 inactivation increased substantially the incidence of early lesions in Ptc1 þ/À mice, suggesting an important role for PARP-1 in suppression of early cerebellar abnormalities. Because the brain is an organ constantly exposed to oxidative stress and damage, it is possible that lack of PARP-1 activity in GCPs may Representative skin from a PARP-1 À/À /Ptc1 þ/À mouse showing three independent microscopic BCC-like tumors in the same microscopic field (700 lm). (D) A 3.8-fold increase in tumor multiplicity ( Ã P 5 0.0031) was observed in PARP-1 À/À /Ptc1 þ/À mice compared with PARP-1 þ/þ /Ptc1 þ/À mice.
M. Tanori et al. lead by itself to genomic instability and hyperproliferation of neural progenitor cells in the developing cerebellum, significantly increasing formation of initial lesions. Consistent with this hypothesis, we detected a significant increase in spontaneous formation of c-H2AX foci in GCPs from PARP-1 À/À /Ptc1 þ/À mice. However, the lack of correlation between frequency of abnormal EGL proliferation areas and development of full CNS malignancy in PARP-1-null mice suggests that, without additional genetic damage (e.g. exogenous damage by radiation), the vast majority of abnormal hyperplastic regions undergo regression, possibly by differentiation or apoptotic processes.
Mice lacking one Ptc1 allele develop a high incidence of medulloblastoma (up to 80%) (27) following radiation damage in neonatal cerebellum. We found that PARP-1 inactivation further increased tumorigenesis in a setting of Ptc1 heterozygosity. In fact, 100% of irradiated PARP À/À /Ptc1 þ/À in this study developed aggressive cerebellar tumors. This strongly suggests a cooperation of DNA DSB processing and deregulated sonic hedgehog signaling in neuronal cells.
In mice lacking one functional Ptc1 copy, the major pathway to medulloblastoma is thought to involve loss of the remaining normal Ptc1 allele. Actually, Ptc1 loss of heterozygosity can be detected in early preneoplastic cerebellar lesions, and time course studies suggest a steadily increase of loss of heterozygosity rate during medulloblastoma tumorigenesis in Ptc1 þ/À mice (28). Concordantly, CGH profiles indicate that medulloblastomas from PARP/Ptc1 mutants show characteristic losses of genetic material around the Ptc1 locus, similar to medulloblastomas from Ptc1 þ/À mice. Mechanistically, this suggests that PARP-1 deficiency may promote tumorigenesis in irradiated Ptc1 þ/À mice by facilitating loss of the remaining normal Ptc1 allele, consistent with the observation that inefficient damage signaling and repair of DNA breaks leads to enhanced chromosomal rearrangements (39). Interestingly, medulloblastomas arising in a variety of mouse models with combinations of targeted deletions in DNA damage signaling and repair genes, such as PARP-1/p53, XRCCA/p53, p53/ p18 Ink4c and Brca2 Nestin-cre /p53, also show Ptc1 loss, suggesting that the major pathway to medulloblastoma involves loss of this gene (31,(40)(41)(42).
Our findings identify DNA break processing as a critical factor in cerebellum tumorigenesis. This view is supported by the observation that PARP-1 inactivation affects DNA damage recognition through phosphorylation signaling, by enhancing expression of the DNA DSB marker histone c-H2AX and increasing phosphorylation levels of Ser18-p53. Consistently, we found longer persistence of c-H2AX foci in PARP-1 À/À cells in response to genotoxic stress, reflecting defects in DNA damage repair. Whether this is due to unresolved SSBs persisting into the S phase of the cell cycle and collapsing replication forks to form DSBs (43) or, as recent findings suggest, to a direct contribution of PARP-1 to DSB repair remains to be determined (44); however, PARP-1 is clearly important for maintenance of DNA integrity in proliferating GCPs after genotoxic stress.
The role of PARP-1 in the apoptotic process is complex and remains to be determined as conflicting data showing inhibition, lack of effect or increased apoptosis upon PARP-1 chemical or genetic abrogation have been reported in different cell types and tissues after genotoxic stress. Particularly, PARP-1-deficient neuronal cells were reported to be resistant to ischemia or neurotoxic cell death (45). In this study, we show significantly increased apoptotic response to radiation in PARP-1 À/À /Ptc1 þ/À compared with PARP-1 þ/þ /Ptc1 þ/À EGL. This is consistent with increased phosphorylation levels of Ser18-p53 in irradiated PARP-1 À/À cerebella, triggering p53dependent apoptosis and cell cycle arrest. This is also in keeping with earlier studies showing that PARP-1 is an essential survival factor following genotoxic exposures, allowing cells to receive appropriate signals for efficient DNA repair and avoiding genomic rearrangement, prolonged cell cycle arrest and apoptosis (12).
Taken together, our data on the effect of PARP-1 loss in the cerebellum show that PARP-1 deficiency leads-through unrepaired DNA damage-to increased cell death in GCPs, and subsequent apoptosis, with escaping cells being genetically altered. This may help to explain both the disorganization of the cerebellar cytoarchitecture and the high susceptibility to brain tumor development in irradiated PARP-1 À/À /Ptc1 þ/À mice.
Tumor development in PARP-1/Ptc1 double-mutant mice often as multiple independent tumors, within 30 weeks of age. This is remarkable for BCC that, even in responsive Ptc1 þ/À mice, shows low frequency, generally delayed onset and development as a solitary tumor. Consequent to inactivation of PARP-1, we also observed highly increased incidence and more rapid onset of preneoplastic skin lesions in irradiated Ptc1 þ/À mice. These results show that PARP-1 function is required to prevent skin cancer formation in response to DNA damage. The PARP-1/Ptc1 double-mutant mice generated here represent a unique new model for BCC as inherent resistance of the mouse to BCC induction by chemicals or physical agents has hampered the study of both the pathogenesis and treatment of this very common human tumor.
The significance of PARP-1 in the pathogenesis of human BCC remains to be explored. Because the skin is an organ constantly exposed to environmental stress, it is conceivable that lack of PARP-1 activity in keratinocytes may prevent elimination of cells that contain damaged DNA and may therefore be susceptible to additional events leading to cancer. Although deletions affecting the PARP-1 locus have not been reported in human cancer, numerous studies have shown overall risk modulation of BCC by variant alleles for DNA strand break repair genes, including XRCC1, that is recruited to sites of SSB repair by PARP-1 (47,48).

Conclusions
PARP-1 has been gaining increasing interest as a therapeutic target for many diseases, including cancer, and PARP-1 inhibitors are shown to be effective in specifically killing cells and tumors with DNA repair defects, such as BRCA gene mutations (49). In fact, PARP-1 inhibitors have already entered clinical trials as a promising strategy in oncology. Concerns relating to systemic treatment with PARP-1 inhibitors are the impairment of DNA repair in normal tissues and the high risk of secondary malignancies because of potential mutagenesis/carcinogenesis linked to the inhibition of DNA repair. Our data support the notion that PARP-1 inhibition may potentiate radiation effects by suppressing DNA repair and improving tumor killing. On the other hand, our findings of increased tumor formation in the cerebellum and in the skin of Ptc1 þ/À mice after exposure to radiation damage raise concern as to increased risk for tumor development since PARP-1 inhibition may unmask recessive mutations in tumor suppressor genes. In line with our results, a recent study has identified PARP-1 protein as a critical factor in cancer susceptibility, revealing an important role of PARP-1 in suppressing mammary tumorigenesis in mice (50). While the clinical benefit of PARP-1 inhibitors is being tested, the inherent risks of long-term exposure in normal tissue will certainly warrant additional studies.