PARP1 Deficiency Reduces Tumour Growth by Decreasing E2F1 Hyperactivation: A Novel Mechanism in the Treatment of Cancer

Simple Summary PARP1 is one of the best characterized enzymes in DNA repair and an attractive target for drug design in cancer therapies. Beyond repair, PARP1 modulates the activity of several transcription factors, of which E2F1 stands out given its critical role in cell cycle regulation. Here, we show that the aberrant activation of E2F1 found in several types of cancer can be alleviated by inactivating PARP1. We believe our findings show the potential of designing novel protein-protein disruptors able to reduce the oncogenic activity of E2F1. Abstract In recent years, poly (ADP-ribose) polymerase (PARP) inhibitors have been evaluated for treating homologous recombination-deficient tumours, taking advantage of synthetic lethality. However, increasing evidence indicates that PARP1 exert several cellular functions unrelated with their role on DNA repair, including function as a co-activator of transcription through protein-protein interaction with E2F1. Since the RB/E2F1 pathway is among the most frequently mutated in many tumour types, we investigated whether the absence of PARP activity could counteract the consequences of E2F1 hyperactivation. Our results demonstrate that genetic ablation of Parp1 extends the survival of Rb-null embryos, while genetic inactivation of Parp1 results in reduced development of pRb-dependent tumours. Our results demonstrate that PARP1 plays a key role as a transcriptional co-activator of the transcription factor E2F1, an important component of the cell cycle regulation. Considering that most oncogenic processes are associated with cell cycle deregulation, the disruption of this PARP1-E2F1 interaction could provide a new therapeutic target of great interest and a wide spectrum of indications.


Introduction
The poly (adenosine diphosphate (ADP)-ribose) polymerase (PARP) family of proteins comprises 18 members in mammals. Among them, the founding member, PARP1, is best known by its ability to attach polymers of ADP-ribose units onto acceptor proteins using NAD + as substrate [1,2]. The best-known function of poly(ADP-ribosyl)ation (PARylation) is its role as a key regulator of

Deletion of Parp1 Rescues Embryonic Lethality in Rb-Deficient Mice
To study the functional interactions between PARP1 and the RB/E2F pathway, we first investigated the effect of the absence of Parp1 on the development of Rb-null embryos. It is well established that the lack of RB has dramatic consequences on embryonic development. These include impairment of erythroid and neuronal differentiation accompanied by several extra-embryonic abnormalities, resulting in early lethality [11][12][13][14]. Many of these features are also observed when overexpression of E2F1 occurs; in fact, the elimination of Rb function results in dysregulation of E2f1 activity, while the survival of homozygous Rb mutant embryos extends until E17 in the absence of E2F1 [15]. Therefore, Rb-null embryos provide a remarkable model for the study of the role of PARP1 in the regulation of E2F1 function. If PARP1 increases the transcriptional activity of E2F1, deletion of Parp1 should be able to counteract the consequences of the absence of Rb and, as a result, the lifespan of Rb-null embryos should be extended. In keeping with this hypothesis, the survival of Parp1 −/− Rb −/− embryos was extended to E16.5, while viability of Parp1 +/+ Rb −/− embryos steeply declined around E13.5, with no viable embryos observed beyond E14.5 ( Figure 1A). Strikingly, this effect in the lethality of Parp1 −/− Rb −/− mice resembles the one reported in E2f1/Rb1 double-null mutants [15]. Parp1 −/− Rb −/− mice also display large superficial vessels, easily observed in the vasculature of the head, which are similar to those present in Parp1 +/+ Rb +/+ embryos ( Figure 1A). A clear reduction in apoptosis was observed in both liver and placenta ( Figure 1B); however, these organs retained high proliferation rates, as demonstrated by mitosis-specific phosphorylated histone H3 labelling ( Figure 1B). Curiously, peripheral blood smears obtained from Parp1 +/+ Rb +/+ and Parp1 −/− Rb +/+ revealed that the absence of Parp1 seems to have a negative effect on the number of enucleated erythrocytes. However, Parp1 −/− Rb −/− embryos display a phenotype similar to the one of wild type embryos, which has also been reported for E2f1 −/− Rb −/− mutants [15], thus demonstrating that failure of erythrocyte maturation can be partially rescued by Parp1 deficiency ( Figure 1C). All in all, these findings reveal the impact of Parp1 loss on the regulation of the RB/E2F pathway, suggesting PARP1 as a positive transcriptional modulator of E2F1. Furthermore, to our knowledge, this is the first demonstration that embryonic lethality induced by RB loss can be delayed by an E2F1 activity-reducing compensatory gene loss.

Deficiency of PARP1 Reduces Tumour Growth in Vitro and in Vivo
Since deregulation of the RB pathway is shared by most human malignancies, we next investigated whether PARP1 may be also involved in the regulation of the RB/E2F pathway in cancer cells. To test this possibility, we took advantage of two different but complementary models. We first used a well-known mouse model of pituitary tumours based on Rb +/mice [12,16]; these mice developed tumours in the intermediate lobe of the pituitary with high penetrance but a long latency period, implying that additional mutagenic 'hits' are necessary for cell transformation. In keeping with this, 92.3% (12/13) of Parp1 +/+ Rb +/− mice developed pituitary tumours by 8 months of age. In contrast, tumour development was less frequent in Parp1 −/− Rb +/− mice (77.8%; 14/18). In addition to reducing tumour penetrance, the absence of PARP1 was also associated with smaller tumour volume ( Figure 2). To further explore the role of the interaction between PARP1 and the RB/E2F pathway in oncogenesis, we extended our study to the gliomagenesis model that was described in our previous study [17]. As this model combines the over-activation of Ras signalling pathways and the deletion of the tumour suppressor pRb, it allows testing the role of PARP1 in a context where hyper-activation of E2f1 function is associated with malignant progression. As Figure 3A depicts, Parp1 +/+ cRb −/− Ras V12 glial cells displayed a transformed phenotype, characterized by the presence of prominent morphological changes and an increased proliferative rate compared with control cells. Immunoblot analysis of protein expression of both groups revealed a slight increase of activated p53 in Parp1 −/− cells, but similar levels of phosphorylated γH2AX ( Figure S1). As expected, this transformed phenotype was partially reversed in Parp1 −/− cRb −/− Ras V12 cells, thus indicating that the absence of Parp1 can ameliorate the outcome of cells with a strong activation of the E2f1 transcriptional activity. Finally, to investigate whether this effect could also be observed in tumour progression in vivo, we used a preclinical glioma model generated by inoculation of either Parp1 +/+ cRb −/− Ras V12 or Parp1 −/− cRb −/− Ras V12 glial cells in mice. Not surprisingly, all mice transplanted with Parp1 +/+ cRb −/− Ras V12 cells developed tumours, while tumours were detected in only 3 out of 12 mice injected with Parp1 −/− cRb −/− Ras V12 cells. To compare tumours from both cohorts, we defined the end-point when the tumour reached a diameter of 12 mm. The Kaplan-Meier shows a significant retardation of Parp1 −/− tumours compared to the control group. Besides, the volumes of tumours in the Parp1 +/+ cohort were significantly bigger than those deficient of Parp1 ( Figure 3B). This result reflects the reduced oncogenic potential of Parp1-null cells and highlights the role of Parp1 as a coactivator of E2f1. When tumour pathology was assessed, Parp1 −/− cRb −/− Ras V12 tumours showed lower phospho-histone H3 staining than Parp1 +/+ cRb −/− Ras V12 tumours ( Figure 3B), a finding that suggests a decreased proliferation rate in the former. Additionally, caspase-3 activity increased in Parp1 −/− cRb −/− Ras V12 tumours ( Figure 3B), which may reflect the reduced tumorigenic potential of these cells. Altogether, these findings demonstrate that the lack of PARP1 reduces cell proliferation in experimental models characterized by the existence of hyperactivation of E2F1 transcriptional activity.

PARP1 Deficiency Inhibits E2F1 Transcriptional Activity
We next investigated whether PARP1 can directly regulate E2F1 transcriptional activity [18,19]. We first looked at how this interaction takes place in an in vitro setting. The affinity constant (K D = 2.2 × 10 −5 M) measured by biolayer interferometry suggests that this molecular interaction leads to the formation of transient and non-obligate complexes, with a moderate affinity, as with most of protein-protein interactions, which would be in line with the notion that PARP1 acts as a co-transcriptional activator at the beginning of the phase S ( Figure 4A). This is further supported by the co-localization of PARP1 and E2F1 in G1/S-phase cells after serum starvation. As shown in Figure 4B and Figure S2, this interaction tends to take place shortly after cells are released from starvation, peaking 4 h after incubation with fresh medium. Furthermore, PARP1-E2F1 interaction occurs on specific response elements that are known to be located on regulatory regions of E2F1-response genes [20][21][22], such as Cyclin-A and E2F1 ( Figure 4C). Taken together, these data suggest that the interaction between PARP1 and E2F1 takes place at the beginning of the S-phase at regulatory regions of E2F1-target genes which contain the E2F1 consensus sequence. To complement this data, Parp1 +/+ Rb +/+ , Parp1 −/− , Parp1 +/+ Rb −/− , Parp1 −/− Rb −/− , and Parp1 +/+ fibroblasts were transfected with a reporter construct driven by a minimal promoter with upstream E2F binding sites (pE2F-Luc) [23]. As it was previously reported [24], Rb-deficient cells showed a sharp increase in luciferase activity in comparison to Rb-expressing cells. However, this hyper-activation of E2f1 was counteracted by the concomitant loss of Rb and Parp1, thus demonstrating the capacity of Parp1 to regulate the transcriptional activity of E2f1 in the absence of pRb ( Figure 4D). Furthermore, Parp1 -/fibroblasts exhibited a decreased EdU incorporation as compared to Parp1 +/+ cells ( Figure 4E), while the lack of Parp1 did not induce senescence nor apoptosis in these cells ( Figure S3). Altogether, these findings indicate that Parp1 modulates the E2f1 transcriptional activity and this activity correlates with the regulation of cell proliferation. In summary, we propose a model in which PARP1 displays transcriptional co-activator activity through interaction with E2F1 on DNA ( Figure 4F). Step correction: Start of association, Start of dissociation

Discussion
E2F activity is considered critical to cell cycle control and its deregulation results in aberrant cell proliferation, whose consequences are best demonstrated during embryonic development and in tumorigenesis. Furthermore, the conserved functions of E2Fs during development suggest that its cancer-related proliferative roles represent a recent evolutionary adaptation, which reinforces the potential of E2F family members as feasible targets for cancer therapy [5]. The results obtained in this study prove that PARP1 modulates E2F1 transcriptional activity, an effect that allows it to play a prominent role in the regulation of proliferation. Furthermore, owing to this, ablation of Parp1 transcriptional activity was shown to counteract the effects of E2f1-induced hyper-replication on embryonic development and tumour growth. This finding suggests a potential for the use of PARP1/E2F1 interaction disruptors as anti-tumour agents. Supporting this possibility, we showed that Parp1 depletion can dramatically reduce the growth of tumoral cells, both in vitro and in vivo. In this regard, we cannot completely rule out an impaired repair ability of the cell due to Parp1 deletion contributing to this effect. In fact, a faulty DNA repair could explain the greater number of apoptotic cells observed in Parp1-deficient tumours ( Figure 3B). However, this increased apoptosis was not observed in transformed Parp1 −/− astrocytes in vitro and the protein levels of phospho-p53 Ser15 and phospho-H2AX Ser139 were not significantly different between both groups ( Figures S1 and S3). Additionally, the fact that the E2F1-PARP1 interaction does not seem to rely on the catalytic activity of PARP1 [10] suggests that the cross talk between both proteins might be independent of the enzymatic and DNA binding activity of PARP1, as in the case with NF-κB [25]. Nonetheless, this provides a basis for future experiments involving PARP1/2 enzymatic inhibitors that might clarify the underlying mechanisms in the PARP1-E2F1 interaction. Finally, we propose a model ( Figure 4F), where under physiological conditions, PARP1 acts as a transcriptional co-activator, thus increasing the transcription of E2F1-dependent genes. In the case of a loss of pRB function, the lack of the inhibitory effect of pRB on E2F1-driven transcription results in increased E2F1 activity, causing deregulation of the cell cycle. The concomitant loss of PARP-E2F1 interaction alleviates the lack of pRB, thus reducing the transcriptional activity of E2F1 and ameliorating the cell cycle regulation. Considering that the majority of oncogenic processes are associated with cell cycle deregulation, the disruption of this protein-protein interaction would provide a new therapeutic approach of great interest and a wide spectrum of indications. All in all, the understanding of the biological implications of the interaction between PARP1 and E2F1 could serve as the foundation of novel therapeutic avenues for managing cell cycle deregulation, a hallmark of cancer.

Affinity Measurement
Binding kinetics and K D values were obtained using bio-layer interferometry (FortéBio, Fremont, CA, USA). Recombinant GST-hE2F1 was purified as described elsewhere [27]. GST-E2F1 was immobilized on GST biosensors kindly provided by the manufacturer. The affinity of PARP1 (Trevigen, Burlington, MA, USA) was analysed using serial dilutions on a FortéBio BLItz instrument, using global fitting and the BLItz Pro-1.2.0.49 software (FortéBio, Fremont, CA, USA).

Co-Localization Studies
For co-localization studies, 5 × 10 4 per cm 2 HEK293 cells were seeded on EZ-multiwell slides (Millipore, Burlington, MA, USA). Cells were transfected using equimolar quantities of pEGFP-PARP1 (a gift from A. Chiarugi) and pRFP-E2F1 (a gift from B. Su) and synchronized by double thymidine treatment. Upon release from cell cycle block, cells were fixed with 2% paraformaldehyde (pH 7.4) at various times. Nuclei were stained using DAPI (Invitrogen, Waltham, MA, USA) and fluorescent images from three different experiments were taken using a Leica TCS SP2 microscope.

Incorporation of 5-Ethynyl-2'-Deoxyuridine (EdU)
Mouse fibroblasts were seeded at 5000 cells per cm 2 in 24-well plates. Culture medium was removed twenty-four hours later and replaced with low-serum medium (0.5% FCS). After 48 h, the starvation medium was replaced with high-serum medium (15% FCS) for 16 h and subsequently treated with 10 µM 5-ethynyl-2'-deoxyuridine (EdU) (Sigma-Aldrich) for two additional hours. Nuclei were stained using Hoechst 3358 (Sigma-Aldrich) and fluorescence images were collected from three different experiments.

Luciferase Assays
MEF were seeded in 12-well plates at 2500 cells per cm 2 . Cells were subsequently transfected with vectors pE2F-Luc (a gift from M. Collado) and pCMV-β-Gal (Clontech, Mountain View, CA, USA). The data from three independent experiments were normalized using beta-galactosidase activity.

Animal Studies
Xenografts were established in SCID (Severe Combined Immunodeficiency) mice aged 10 to 12 weeks. Cell implantation was carried out by subcutaneous injection in the hindquarters with 3 × 10 6 transduced astrocytes resuspended in 100 µL PBS 1×. All tumours included in the analysis reached a minimum diameter of 4 mm and mice were euthanized when they approached a maximum diameter of 12 mm. Tumours were considered ellipsoid in shape and their volume was calculated using the equation volume = 0.5 × (length × width) [29]. Genetically modified strains for Rb1, specifically Rb1 knock-out mice [11] and cRb [30], were obtained from the Mouse Models of Human Cancer Consortium (MMHHC) repository. Parp1 −/− strain [31] was a gift from Prof. de Murcia. All animal procedures were approved and performed according to the guidelines set out by the Institutional Ethics Committee for Animal Experimentation.

Immunohistochemistry
Analysis of foetal erythrocytes from cord blood samples was carried out by Wright-Giemsa staining. For immunohistochemical analysis, anti-cleaved caspase 3 (monoclonal, Cell Signaling) and phospho-histone H3 (polyclonal, Cell Signaling Technology, Inc) were used as primary antibodies. Immunohistochemical analysis was performed using a universal second antibody kit that uses a peroxidase-conjugated labelled dextran polymer (Envision Plus, Dako, Glostrup, Denmark). The following primary antibodies were used: anti-cleaved caspase 3 (monoclonal, Cell Signaling Technology, Inc) and phospho-histone H3 (polyclonal, Cell Signaling Technology, Inc).

Statistics
Statistical analysis was performed with ANOVA and Student's t-test for multiple or simple comparisons, respectively. Tukey and Student-Neumel-Kaus tests were used for post-hoc analysis of ANOVA results. Mantel-Cox test was used to analyse Kaplan-Meier curves. In all cases, statistical significance was established at p < 0.05.

Study Approval
All animal procedures were approved and performed according to the guidelines set out by the USC Bioethics Board (protocol No 15005AE/07/FUN01/FIS02/JACP1).

Conclusions
In addition to its role on DNA repair, PARP1 is a co-transcriptional activator of E2F1, thus regulating the progression through the cell cycle. Decreasing PARP1-E2F1 interactions results in reduced transcriptional activity, thus providing of a novel therapeutic avenue for managing cell cycle deregulation, a hallmark of malignant tumours.