Inactive PARP1 causes embryonic lethality and genome instability in a dominant-negative manner

Significance

PARP1 (poly-ADP ribose polymerase 1) is the primary target of PARP enzymatic inhibitors. Therapeutic use of PARP inhibitors for cancer is based on not only the extreme sensitivity of BRCA1/2-deficient cancer cells to PARP1 inhibition but also the nonessential role of PARP1 in normal tissues. Here, we show that in contrast to the normal development of Parp1-null mice, the mouse model heterozygously expressing the catalytically inactive Parp1 (E988A, Parp1^+/A) dies embryonically with high levels of genomic instability. The results reveal the severe dominant-negative impact of catalytically inactive PARP1, indicating the presence of enzymatically inactive PARP1 is much more damaging to normal tissues than previously anticipated. These findings provide a mechanism for clinical PARP inhibitors’ unexpected normal tissue toxicity.

Abstract

PARP1 (poly-ADP ribose polymerase 1) is recruited and activated by DNA strand breaks, catalyzing the generation of poly-ADP-ribose (PAR) chains from NAD+. PAR relaxes chromatin and recruits other DNA repair factors, including XRCC1 and DNA Ligase 3, to maintain genomic stability. Here we show that, in contrast to the normal development of Parp1-null mice, heterozygous expression of catalytically inactive Parp1 (E988A, Parp1^+/A) acts in a dominant-negative manner to disrupt murine embryogenesis. As such, all the surviving F1 Parp1^+/A mice are chimeras with mixed Parp1^+/AN (neoR retention) cells that act similarly to Parp1^+/−. Pure F2 Parp1^+/A embryos were found at Mendelian ratios at the E3.5 blastocyst stage but died before E9.5. Compared to Parp1^−/− cells, genotype and expression-validated pure Parp1^+/A cells retain significant ADP-ribosylation and PARylation activities but accumulate markedly higher levels of sister chromatid exchange and mitotic bridges. Despite proficiency for homologous recombination and nonhomologous end-joining measured by reporter assays and supported by normal lymphocyte and germ cell development, Parp1^+/A cells are hypersensitive to base damages, radiation, and Topoisomerase I and II inhibition. The sensitivity of Parp1^+/A cells to base damages and Topo inhibitors exceed Parp1^−/− controls. The findings show that the enzymatically inactive PARP1 dominant negatively blocks DNA repair in selective pathways beyond wild-type PARP1 and establishes a crucial physiological difference between PARP1 inactivation vs. deletion. As a result, the expression of enzymatically inactive PARP1 from one allele is sufficient to abrogate murine embryonic development, providing a mechanism for the on-target side effect of PARP inhibitors used for cancer therapy.

PARP1 belongs to the family of poly-ADP ribose polymerases (PARPs) that share a conserved ADP-ribose (ADPr) transferase (ART) domain. PARP1 and the related PARP2 are recruited to DNA strand breaks, where they are activated to catalyze the transfer of the ADP-ribose to the protein substrate (mono-ADP-ribosylation, MARylation) and then chain extension (Poly-ADP-ribosylation, PARylation) to form poly-ADP-ribose (PAR). PAR promotes chromatin relaxation and recruits other repair proteins, including the XRCC1-Ligase3 complex that ligates single-strand DNA nicks, such as those generated during base excision repair (BER) (1–3). PARP1 is more abundant than PARP2, has a higher affinity for diverse DNA lesions, and accounts for >80% of DNA damage-induced PARP activity (4). Yet, Parp1-null mice are viable, fertile, and of standard size (5). Indeed, the use of PARP enzymatic inhibitors (PARPi) for cancer therapy is based not only on the extreme sensitivity of BRCA1- or BRCA2-deficient cancer cells to PARP1 loss, but also on the nonessential role of PARP1 in normal tissues (6, 7).

In addition to blocking its enzymatic activity, PARPi also induces a phenotype termed “PARP trapping”, characterized by the persistence of damage-induced PARP1/2 foci and prolonged retention of PARP1/2 on damaged chromatin (8–11). Loss of PARP1, but not PARP2, causes marked resistance to PARP inhibitors (10, 11), highlighting the unique importance of PARP1 inhibition in cancer therapy. Several mechanisms have been proposed to explain PARP1 trapping (12). In vitro, PARPi prevents PARP1 auto-PARylation, which correlates with the release of PARP1 from DNA ends (9). But this mechanism alone cannot explain why PARPi with similar IC50s for enzymatic inhibition have different abilities to “trap” PARP1. Structural analyses later showed that the nonhydrolyzable NAD+ analog benzamide adenine dinucleotide could reverse-allosterically enhance PARP1 affinity for DNA ends (13), providing an explanation for the differential trapping by inhibitors with distinct chemical structures. PARPi attenuates PAR, which recruits XRCC1-LIG3 and other repair complexes (11). Using quantitative live-cell imaging, we recently showed that PARP1 molecules exchange rapidly at microirradiation sites in cells (11), even in the presence of clinically effective PARPi, suggesting that the PARPi-induced persistence of PARP1 foci may reflect continual recruitment of different PARP1 molecules to the unrepaired DNA lesion due to delayed repair. In this regard, the release and exchange of PARP1 from model DNA ends in vitro can also occur independently of auto-PARylation and in the presence of clinically relevant PARPi (12, 14–16).

Although the therapeutic effects of PARP inhibition and trapping on tumor cells have been studied extensively, less is known about the impact of inactive PARP1 on normal cells, despite accumulating evidence of clinical PARPi toxicities in patients as hematological and gastric-intestinal toxicities, and inhibitor-treated mouse models as acute hematological toxicities (17–20). To address this question and understand the physiological impact of inactive PARP1 protein, we introduced a catalytically inactive point mutation into the endogenous Parp1 locus. Here we show that, in contrast to the normal development of Parp1-null mice, heterozygous expression of catalytically inactive Parp1 (E988A, Parp1^+/A) acts in a dominant-negative manner to disrupt embryogenesis and genome stability. Parp1^+/A cells are characterized by the elevated formation of mitotic bridges and sister chromatid exchanges (SCEs). The genome toxicity is due to the accumulation of inactive PARP1 protein at DNA damage sites, but not the loss of ADP-ribosylation and PARylation activity. Mechanistically, we show that inactive PARP1-E988A is recruited to DNA damage sites where it forms persistent foci. Especially, the inactive Parp1 blocks the resolution of DNA damage induced by alkylating agents and topoisomerase inhibitors beyond the loss of PARP1, but not clean DNA double-strand break (DSB) repair via homologous recombination (HR) or nonhomologous end-joining (NHEJ), measured by lymphocyte development, meiosis, and reporter assays. The results demonstrate a clear difference between PARP1 inhibition vs. PARP1 deletion on normal tissues, providing the molecular mechanism for the initially unexpected and significant adverse effects of PARPi.

Results

The Generation and Characterization of Catalytically Inactive Parp1 Murine Model.

To address the impact of inactive PARP1 on normal issue, we generated the Parp1^AN allele by introducing the catalytically inactive E988A missense mutation (21) and an FRT-NeoR-FRT cassette into the endogenous Parp1 locus of mice (Fig. 1 A and B). The E988A mutation abrogates the PARylation activity of PARP1 without affecting its DNA binding (21) or MARylation activity at high NAD+ concentration (11). Since the FRT-NeoR-FRT cassette blocks transcription (Fig. 1A), Parp1^AN behaves like a null allele and, as anticipated, Parp1^+/AN pups were born at the expected Mendelian ratio (Fig. 1C). To remove the cassette and generate the desired Parp1^A allele, Parp1^+/AN mice were bred with Rosa^FLIP/FLIP mice that constitutively expresses the FLIPase from the blastocyte stage (E3.5) in widespread tissues including the reproductive organs (Jax Strain No. 003946 in 129Sv) (22, 23). Regardless of whether the Rosa^FLIP/FLIP came from the paternal or the maternal side, the F1 Parp1^+/A pups were significantly underrepresented among the progeny (1/6 of expected) (Fig. 1D). They were also ~25% smaller than their littermates at pre-weaning ages (Fig. 1E) and many (~45.8%) had kinked tails (Fig. 1F), a phenotype previously seen in genomic instability mouse models (e.g., H2ax^−/− or Brca1 mutants), but not in Parp1^−/− mice, suggesting that catalytically inactive Parp1 polypeptides can disrupt genome integrity in a dominant-negative manner.

Fig. 1.

Expression of catalytically inactive Parp1 impairs murine development. (A) Mouse Parp1 E988 E->A targeting strategy. The schematic diagram represents the murine Parp1 germline structure (top row), targeting vector (second row), targeted allele (Parp1^AN, third row), and the recombined E988A expression allele without neo (Parp1^A, bottom row). The presence of the NeoR in the Parp1^AN allele interferes with the expression of Parp1, resulting in no or little Parp1 expression. Map not shown to scale. (B) Southern blot analysis of targeted Parp1 ^+/AN ES cell DNA using 5′ and 3′ probes. Left: wild-type ES cells. Right: targeted ES cells. Top band: germline. Lower band: targeted. (C) The birth rate of F1 pups from Parp1^+/AN x Parp1^+/+ crossing. P-values were calculated via the χ² test. (D) The birth rate of F1 pups from Parp1^+/AN x Rosa^Flip/Flip crossing. P-values were calculated via the χ² test. (E) Body weight of day 7 to 10 Rosa^+/Flip Parp1^+/A and age-matched control Rosa^+/Flip Parp1^+/+ littermate pups. The Rosa^+/Flip Parp1^+/A pups were consistently smaller in this age window. P-value was calculated using multiple student’s t test. **P < 0.01, ***P < 0.001. (F) Representative images and frequency of F1 Parp1^+/A chimeras or Parp1^+/+ mice with kinked tails.

F2 Parp1^+/A Mice Die during Early Embryonic Development.

Curiously, all the F2 pups (n = 52) of mating between F1 Parp1^+/A mice and WT partners (both genders) were Parp1^+/+, regardless of Tp53 status(Fig. 2A, P < 3.9 × 10⁻⁹) and despite equal representation of Parp1^E988A and Parp1⁺ alleles in the sperm of F1 Parp1^+/A mice (Fig. 2B). Timed mating showed that F2 Parp1^+/A embryos were recovered with normal early-stage blastocyst morphology at E3.5 (Fig. 2C) but could not be recovered at E9.5, indicating F2 Parp1^+/A animals likely suffered early embryonic lethality (Fig. 2A). Moreover, despite the known requirement for HR in germ cell development and meiosis, testis architecture (Fig. 2D), sperm counts (Fig. 2E), testis weight (Fig. 2F), and sperm mobility were normal in young F1 Parp1^+/A mice used for breeding (SI Appendix, Fig. S1A). The lack of F2 Parp^+/A mice was perplexing given the viability of F1 Parp1^+/A pups (Fig. 1D). We hypothesized that the surviving F1 Parp1^+/A mice might be chimeras, which carry Parp1^+/AN cells that had escaped FLIPase excision and whose presence, although low in frequency, support the embryogenesis of F1 Parp1^+/A mice. To test this possibility, we regenotyped F1 Parp1^+/A mice by PCR using both the standard triplex primers (detecting AN, A, and + alleles) as well as primers that only detect the AN allele (SI Appendix, Fig. S1B). All F1 Parp1^+/A mouse tail DNAs tested AN-positive (allele frequency: 1% to 10%) in the AN-specific PCR assay (Fig. 2 G and H and SI Appendix, Fig. S1 C and D). We note that the PCR product for the A allele shares 230-bp sequence identical to that of the + allele, except the ~80-bp insertion corresponding to the FRT site and its surrounding sequence (SI Appendix, Fig. S1B). As a result, when both A and + products are present (e.g., in Parp1^+/A), the two PCR products sometimes form a heteroduplex that migrates between the WT and A PCR product (Fig. 2G). Moreover, qPCR analyses of peripheral blood from F1 Parp1^+/A mice at different ages showed a progressive increase in AN allele frequency from ~1% at 3 wk to ~100% at 64 wk (Fig. 2I). This is consistent with the low chimerism in the tail DNA collected at 7 d after birth (Fig. 2 G and H and SI Appendix, Fig. S1 C and D). Thus, we concluded that pure heterozygosity abrogates the embryonic development of Parp1^+/A mice.

Fig. 2.

Heterozygous expression of inactive Parp1-E988A causes embryonic lethality in mice. (A) The frequency of F1 Parp1^+/A pups from Parp1^+/A x Parp1^+/+ crossing with or without Tp53 deficiency at different developmental stages. P-values were calculated via the χ² test. (B) PCR identification of Parp1^Aallele in F1 Parp1^+/A mice sperm. (C) Morphology of F2 Parp1^+/+ and Parp1^+/A early-stage blastocysts at E3.5. (D) Histology with H&E staining showing spermatogenesis in young (<8 wk) male Parp1^+/+, Parp1^+/−, and F1 Parp1^+/Atestes. (E) Epididymis mature sperm count of Parp1^+/+, Parp1^+/−, and Parp1^+/A adult males. The P-value was calculated based on one-way ANOVA. ns: P > 0.05. (F) Testes weight from F1 Parp1^+/Aand control male Parp1^+/+ and Parp1^+/− mice. (G and H) Standard triplex genotyping PCR (G) and Targeted AN-specific PCR (H) of tail DNA from 7 to 10-d old F1 Parp1^+/A chimera mice and controls. ID number 1 to 3 identifies unique pups. The lower band of the doublet in panel G represents a hybrid between the A and WT allele PCR products due to their share 230-bp sequence. This band only appears in PCR from Parp1^+/A mice when both the WT and the A alleles are present. (I) qPCR of Parp1^AN in peripheral blood of F1 Parp1^+/A mice of different ages.

Normal Lymphocyte Development and Spermatogenesis in Viable Young (<8 wk) F1 Parp1^+/A Chimeric Mice.

PARP1 is activated by DNA breaks. To understand the mechanism of growth retardation, we then asked whether inactive Parp1 blocks DNA DSB repair. Mammalian cells have two major DSB repair pathways –NHEJ and HR. Specifically, lymphocyte development requires the ordered assembly and subsequent modification of immunoglobulin (Ig) genes through two programmed DSB events–V(D)J recombination and Ig class switch recombination (CSR) (24). Defects in the NHEJ pathway abrogate V(D)J recombination and significantly impair CSR (24), serving as a physiological readout of NHEJ. The Parp1^+/AN chimerism in the F1 Parp1^+/A mice is consistently <10% in peripheral blood by 15 wk of age. So, we focused our analyses of the F1 Chimeric Parp1^+/A on young mice only. Like in Parp1 null mice (5), B and T lymphocyte development measured by both relative frequency of immature and mature cells and developmental state-specific cellularity, were both normal in young F1 Parp1^+/A mice (<8 wk) (Fig. 3 A–C). Representative genotype verified that AN allele frequency in the lymphoid organs is <=10% at the time of analyses (SI Appendix, Fig. S1E). Furthermore, IgH CSR, a sensitivity measure for NHEJ (25), was also normal in the splenocytes purified from young F1 Parp1^+/A mice (<8 wk) (Fig. 3 A and D). Since CSR efficiency can be influenced by cell proliferation, we further measured CSR by gating cells with the same cell division. No measurable defects in CSR were noted in Parp1^+/A splenocytes even after being controlled for cell division (SI Appendix, Fig. S2A). Given the normal spermatogenesis, a process that requires HR, in young Parp1^+/A chimeras, these data together suggest that heterozygous expression of inactive PARP1 does not have a major impact on NHEJ or HR. However, due to the chimerism, although low, we would not be able to rule out minor impacts on HR and NHEJ in the Parp1^+/A mice based on the apparent normal development in F1 chimeric Parp1^+/A mice.

Fig. 3.

Young F1 chimeric Parp1^+/A mice have normal lymphocyte development. (A–D) Representative flow cytometry analyses (A), relative percentage (B), and absolute cell count (C) of Parp1^+/A mice lymphocyte and myeloid development, as well as naïve B lymphocytes CSR efficiency (D). The Student’s t test was applied to calculate the P value. ns: P > 0.05.

Parp1-E988A Forms Persistent Foci at Microradiation Sites without Exchange Defects.

To ascertain whether E988A extends the appearance of PARP1 at DNA damage sites, we measured the kinetics of foci formation by GFP-tagged PARP1 following 405-nm microirradiation in PARP1 knockout human osteosarcoma cell line—U2OS previously generated in our lab (11). We chose the PARP1 knockout cells to prevent the endogenous PARP1 from interfering with the dynamics of the GFP-tagged PARP1. U2OS is commonly used as a model for quantitively live cell imaging due to its flat nuclear shape, reliable and high transfection efficiency, and consistent behavior upon microirradiation. Both GFP-PARP1-E988A and GFP-PARP1-WT proteins rapidly formed foci within 1 min of irradiation, suggesting the E988A mutation does not affect DNA binding by PARP1 (Fig. 4 A and B). As expected, cells expressing GFP-PARP1-E988A failed to form bright PAR-dependent XRCC1 foci, consistent with the lack of PARylation activity (Fig. 4 A and C and SI Appendix, Fig. S2B). However, unlike GFP-PARP1-WT foci, which dissolved within ~10 min, at least 35% of GFP-PARP1-E988A foci persisted for 20 min and 25% lasted to 30 min (Fig. 4 A and D). The lack of XRCC1 foci formation and the persistent PARP1 foci in cells expressing PARP1-E988A are comparable to PARP inhibitor—niraparib-treated cells expressing PARP1 WT (Fig. 4 A–D). The presence of PARP1 E988A at the DNA ends in later time points could physically block the recruitment of other DNA repair factors, including PARP2 (see below).

Fig. 4.

Inactive Parp1 forms persistent DNA damage-induced foci without exchange defects. Summary of quantitative live-cell imaging following laser microirradiation in the PARP1 knockout U2OS cells. The time (minutes) in this figure indicates the time after 405-nm laser-induced microirradiation. (A) Representative images of GFP-tagged PARP1(in green) and RFP-tagged XRCC1 (in red) after microirradiation. Niraparib (1 μM) was added 1 h before the microirradiation. (B) Quantification of relative PARP1 foci intensity (as a ratio of GFP intensity at the foci vs. nuclear average) at 1 min after microirradiation. (C) Quantification of relative XRCC1 foci intensity at 1 min after microirradiation. (D) The normalized kinetics (as a percentage of the maximal relative intensity) of GFP-tagged E988A-PARP1 or wild-type PARP1 in PARP1-deficient U2OS cells with or without 1 μM niraparib after laser microirradiation. ns: P > 0.05, *P < 0.05, **P < 0.01, and ***P < 0.001. (E and F) Representative FRAP images (E) and recovery curves (F) of E988A-PARP1 and wild-type PARP1. The foci area was bleached by 488-nm laser at 1 min after microirradiation.

Persistent PARP1 foci could be caused by allosteric trapping of the same PARP1 molecule at the DNA breaks or by continuous recruitment of different PARP1 molecules to the breaks due to the lack of repair (e.g., reduced PAR-dependent recruitment of the XRCC1-LIG3 complex) (11). To distinguish these two possibilities, we measured florescence recovery after photobleach (FRAP). Upon photobleaching, GFP-PARP1-E988A foci recovered as efficiently as the WT control (11) (Fig. 4 E and F), suggesting that E988A mutation does NOT affect the exchange of PARP1 at the DNA damage foci and the persistency of PARP1-E988A foci is likely caused by continuous recruitment of different PARP1-E988A proteins to the unrepaired breaks. This is similar to what we and others have reported for clinical PARPi-treated PARP1-WT (11). Thus, these data suggest that inactive PARP1 (E988A) can occupy the DNA breaks for an extended time, where it blocks other repair proteins from accessing the ends.

Isolated Parp1^+/A Cells Support Significant DNA Damage-Induced PARylation.

Next, we isolated Parp1^+/A and Parp1^−/− SV40 antigens immortalized murine embryonic fibroblasts (iMEFs) and confirmed by genotyping (SI Appendix, Fig. S2 C and D and Table S1) to understand how inactive Parp1 dominant negatively blocks embryonic development. We also validated that the A allele is expressed in the Parp1^+/A iMEF cells via RT-PCR. The nucleotides encoding Glutamine 988 reside at the junction between Exon 22 and 23 (SI Appendix, Fig. S2E). The targeting created a new MluI digestion site in the cDNA (but not in genomic DNA). MluI digestion of RT-PCR products derived from Parp1^+/A cells detected two bands of equal intensities, indicating the A allele is expressed at levels comparable to the WT allele (SI Appendix, Fig. S2E). Correspondingly, sequencing of the cDNA derived from Parp1^+/A cells also revealed mixed cDNA at the A->C mutation (leading to Glu to Ala) and silent mutation T->C (Asn-> Asn) (SI Appendix, Fig. S2F). Together these data confirmed that the Parp1^A allele is expressed at a comparable level to the WT allele in the Parp1^+/A iMEFs.

In previous studies, we showed that the majority of PARP2 is recruited to DNA damage sites by the PAR chain generated by PARP1 and that the presence of PARP1 protein prevents DNA-dependent recruitment and activation of PARP2 (26). Given loss of both PARP1 and PARP2 led to embryonic lethality, we hypothesized that the presence of inactive PARP1 might simply abolish the PARylation by both PARP1 and PARP2 by blocking the remaining PARP1 and PARP2 activation. To test this, we first measured PARylation-dependent recruitment of XRCC1 to DNA damage foci in Parp1^+/A cells. The intensity of the RFP-XRCC1 foci in Parp1^+/A iMEFs was higher than Parp1^−/− iMEFs while also lower than Parp1^+/+ iMEFs (Fig. 5A). Next, we measured pan-ADP-ribosylation using western blot. The ADP-ribose has a molecular weight of 559.3157g/mol (=0.559 KDa). Thus, MARylation or adding a few ADP-Ribose units cannot cause measurable molecular weight shifts on western blotting. The major molecular weight shift in Fig. 5B results from extensive PARylation. Both baseline and damage-induced auto-ADP-ribosylation of PARP1 and histones are comparable in Parp1^+/A and control Parp1^+/+ iMEFs and significantly higher than that of Parp1^−/− iMEFs (Fig. 5 B and C). This is consistent with the notion that the E988A mutation preserves some mono-ADP-ribosylation (MARylation) activity (11) and the inactive PARP1 (E988A) does not abrogate the activation of the WT PARP1 in the Parp1^+/A cells. In this context, we have shown that E988A PARP1, like WT PARP1, can exchange at the site of damage efficiently (Fig. 4F), which explains why E988A did not prevent WT PARP1 from being activated by DNA breaks. Consistent with Parp1 (regardless of activity) blocking PARP2 activation, Parp2 auto-ADP-ribosylation is highest in Parp1^−/− cells, and lower in both Parp1^+/+ and Parp1^+/A cells. Next, we used an antibody specific for PARylation (cannot detect MARylation). We found that H₂O₂-induced auto-PARylation in Parp1^+/A cells is comparable with that of Parp1^+/− iMEFs, and much higher than that of Parp1^−/− iMEFs (Fig. 5D). This result is consistent with the intramolecular auto-PARylation mode of PARP1. In both Parp1^+/A and Parp1^+/− iMEFs only one copy of the PARP1 can be auto-PARylated. Collectively, the result suggests that the embryonic lethality of Parp1^+/A mice cannot be explained by the lack of overall ADP-ribosylation or PARylation. Instead, the continuous recruitment and the presence of inactive Parp1 protein at DNA damage lesions might physically block DNA repair by preventing other proteins (e.g., PARP2) from accessing the ends.

Fig. 5.

PARP1 E988A causes increased SCE in a dominant-negative manner without ablating DNA damage-induced PARylation in Parp1^+/A cells. (A) Representative images and 1-min intensity of RFP tagged XRCC1 foci in Parp1^+/+, Parp1^−/−, and Parp1^+/A immortalized MEFs after 405-nm laser microirradiation. (B and C) Representative western blot (B) and quantification of pan-AND-ribosylated PARP1 and PARP2 in Parp1^+/+, Parp1^−/−, and Parp1^+/A immortalized MEFs after 0.2 mg/mL of MMS treatment (C). *P < 0.05 and **P < 0.01, P-value was calculated by Student’s t test. (D) Western blot of PAR in H₂O₂ stimulated Parp1^−/−, Parp1^+/+, Parp1^+/−, and Parp1^+/A immortalized MEFs. (E) Representative images showing normal mitotic cells (Top) or cells with a mitotic bridge (Bottom) and frequency of mitotic bridges in Parp1^+/+, Parp1^−/−, and Parp1^+/A immortalized MEFs. (F) Representative images and statistics of SCE rate of Parp1^+/+, Parp1^+/−, Parp1^−/−, and Parp1^+/A ES cells. The P value was calculated based on one-way ANOVA. ***P < 0.001.

Inactivated PARP1 Protein Selectively Blocks Single-Strand Breaks Repair and Topo-II Lesions.

DNA strand breaks, both single- and DSBs, recruit and activate PARP1. To understand the types of DNA repair events blocked by Parp1^A, we isolated Parp1^+/A and Parp1^−/− embryonic stem (ES) cells and confirmed via PCR that they do not have the Parp1^AN allele (SI Appendix, Fig. S2 C and D). ES cells were chosen because their inherent checkpoint defects would allow us to visualize chromosomal instability in metaphase preparations easily. Mitotic bridge formation was increased in both Parp1^+/A and Parp1^−/− ES cells to similar levels (Fig. 5E). Meanwhile, in comparison to the Parp1^+/+ ES cells, the frequency of sister-chromatid exchanges (SCEs) per chromosome increased ~2.4 fold in Parp1^−/− ES cells, and ~5.5-fold in Parp1^+/A ES cells (Fig. 5F). Despite the genomic instability, Parp1^+/A and Parp1^−/− iMEFs proliferated at similar levels as the matched Parp1^+/+ control (5) (Fig. 6A).

Fig. 6.

PARP1 E988A dominant negatively inhibits selective DNA repair pathways without ablating HR or NHEJ repair. (A) The proliferation rate of Parp1^+/+, Parp1^−/−, and Parp1^+/A immortalized MEFs. (B) Sketch diagram and HR efficiency measured by the frequency of GFP+ cells in Parp1^+/+, Parp1^−/−, and Parp1^+/A immortalized MEFs with Parp1^+/+ + ATMi (KU-55933 15 μM, Selleckhem, S1092) control using DR-GFP reporter. Student’s t test was used to calculated P-value, ns: P > 0.05 and *P < 0.05. (C) Sketch diagram and NHEJ efficiency measured by the frequency of GFP+ cells in Parp1^+/+, Parp1^−/−, and Parp1^+/A immortalized MEFs with Xrcc4^−/− immortalized MEFs control using EJ5 reporter. P-value was calculated by Student’s t test, ns: P > 0.05, *P < 0.05, and ***P < 0.001. (D–G) Sensitivity of Parp1^+/+, Parp1^−/−, and Parp1^+/A immortalized MEFs to MMS (D), CPT (E), etoposide (F), and IR (G). The P-value was calculated via Student’s t test and extra sum-of-square F test using the dose–response model. All dots and error bars represent means and SEs. ns: P > 0.05, *P < 0.05, **P < 0.01, and ***P < 0.001.

Using this genotype and expression validated Parp1^+/A iMEF, we measured HR and NHEJ (and microhomology mediated end-joining -MMEJ) using the DR-GFP and EJ5 reporters, respectively. After the reporter has been integrated into the Parp1^+/A and control iMEF, transient expression of I-SceI induces GFP conversion via HR in DR-GFP reporter containing cells (27) and via NHEJ, and to a lesser extent MMEJ, in EJ5 containing cells (28). Consistent with normal meiosis in the young Parp1^+/A F1 chimeras, Parp1^+/A and Parp1^−/− iMEFs support DR-GFP conversion at levels comparable to or even higher than the Parp1^+/+ iMEFs (Fig. 6B). ATM inhibitor treatment significantly reduced DR-GFP conversion in Parp1^+/+ cells, serving as a control (Fig. 6B) (29–31). In parallel, EJ5 conversion is also very efficient in Parp1^+/A and Parp1^−/− iMEFs (Fig. 6C). The elevated levels of DR-GFP in Parp1^+/A and Parp1^−/− iMEFs are not due to high copies of the substrates in Parp1^+/A and Parp1^−/− iMEFs (SI Appendix, Fig. S3 A and B) but might reflect the increased SCE in Parp1^+/A and Parp1^−/− cells (Fig. 5F). Together with the normal lymphocyte development and meiosis in the young F1 Parp1^+/A mice, the results suggest that inactive PARP1 did not abrogate HR or NHEJ.

To understand what other type of DNA repair was blocked by inactive PARP1, we measured DNA damage sensitivity using Parp1^+/A and control iMEFs. Even in comparison to Parp1^−/− iMEFs, Parp1^+/A iMEFs were much more sensitive to the alkylating agent methyl-methanesulfonate (MMS) (Fig. 6D and SI Appendix, Fig. S3C), the Topoisomerase I inhibitor camptothecin (CPT) (Fig. 6E and SI Appendix, Fig. S3D) and, perhaps unexpectedly, the Topoisomerase II inhibitor etoposide (Fig. 6F). Parp1^−/− cells are also sensitive to MMS and CPT, albeit less than their Parp1^+/A counterpart, consistent with a role of PAR and PAR-mediated recruitment of XRCC1 in BER and Top1-poison removal (Fig. 6 D and E). In this context, XRCC1 deficiency has also been linked with MMS and CPT hypersensitivity (2). But etoposide sensitivity has not been reported in XRCC1-deficient cells. The hypersensitivity of Parp1^+/A cells to Topo-II block and the normal lymphocyte development and spermatogenesis in young F1 Parp1^+/A chimera mice, together with their HR and NHEJ/MMEJ proficiency measured by the DR-GFP and EJ5 reporters (Fig. 6 B and C), suggest inactive PARP1, but not lack of PARP1, might block the processing of end-blocked lesions, especially during replication, likely independent of NHEJ or HR. Consistent with this notion, Parp1^+/A and Parp1^−/− iMEF displayed relatively consistent and comparable levels of hypersensitivity to radiation (Fig.6G and SI Appendix, Fig. S3E).

Discussion

Taken together, the early embryonic lethality of Parp1^+/A mice provides strong evidence for substantial toxicity of catalytically inactive PARP1 protein in otherwise normal tissues. This is in sharp contrast to the normal development of Parp1^−/− mice. In this context, Parp1^+/A cells share many features with PARPi-treated normal cells (18) and XRCC1-deficient cells (2, 3), including BER deficiency, increased SCE, and persistent PARP1 foci without allosteric locking (11, 14). Conversely, unlike the PARPi-treated cells, the overall ADP-ribosylation activity in the Parp1^+/A cells closely resembles the Parp1^+/+ cells, and is much higher than in Parp1^−/− cells (Fig. 5 B and C). The damage-induced auto-PARylation of PARP1 in Parp1^+/A cells is similar to Parp1^+/− cells (Fig. 5D), consistent with both lines having only one copy of active PARP1 that can auto-PARylate itself. Loss of PARP1 leads to resistance to PARP inhibitors (11, 32). Together these data support a model in which the presence of inactive PARP1, but not the lack of PARP1-mediated PARylation in general, underlies the normal tissue toxicity of PARP inhibition. Moreover, it suggests that if half of the PARP1 is inactive, it is sufficient to cause major normal tissue toxicity. Given the development impairment of Parp1^+/A cells even without small molecule inhibitors, and the efficient exchange of PARP1-E988A at the microirradiation sites, our results also suggest that the small molecule induced reverse-allosteric locking of PARP1, although may be important for inhibitor function, are not essential for normal tissue toxicity. It is the presence of inactive PARP1 with high-affinity DNA binding that is toxic. In this context, the kinetics of PARylation, thus enzymatic activity might be important for preventing persistent PARP1 foci (33). HPF1 was recently identified as an essential modulator for PARP1 activity on Serine (34). Lack of HPF1 and similar modulators also causes persistent PARP1 foci and toxicity, offering additional strategies for targeting PARP1.

How does the inactive PARP1 differ from PARP1 deletion? MMS-induced PARP2 auto-ADP-ribosylation increased in Parp1^−/− cells, but not in Parp1^+/A cells (Fig. 5 B and C), suggesting the presence of inactive Parp1 blocks PARP2 activation and automodification. It is possible PARP2 is only one of the many repair proteins that is blocked by inactive PARP1. As a result, while both Parp1^+/A cells and Parp1^−/− cells are hypersensitive to MMS and Topo I inhibitors, the hypersensitivity to Topo II inhibitors seems unique to the presence of inactive PARP1 and has not been found with PARP1 or XRCC1 deletion (2, 3). Among the genotoxic anticancer therapies, alkylating agents and Topo II inhibitors carry a particularly high risk for therapy-induced MDS/AML (t-MDS/AML) and clonal hematopoiesis. Thus, the hypersensitivity of Parp1^+/A cells to these agents might explain the severe hematological toxicity and t-MDS/AML associated with PARP inhibitors (17, 19), a hypothesis that can be tested by somatic inactivation models.

The Parp1^+/A mouse model described here provides strong evidence for severe toxicity of the inactive PARP1 protein during embryonic development in vivo. PARP1 resembles the ATM, ATR, and DNA-PK kinases, which also exhibit more severe phenotypes when enzymatically impaired than when deleted entirely (35). They all share a common feature. All are recruited to and activated at the site of DNA damage. The activation is coupled with a rapid exchange of the proteins at the DNA damage site. The presence of the inactive protein not only blocks the further activation of the signaling cascade, but also occupies the DNA lesion, where it blocks the activation of other sensors or DNA repair itself. This model is consistent with the fact that no “inactivation” mechanism for these DNA damage response factors exists. These insights should guide the development of therapeutic agents that circumvent the normal tissue toxicities associated with current PARP1, ATM, and ATR inhibitors.

Materials and Methods

Mice.

The 5′ and 3′ arms of the Parp1^E988A targeting construct were amplified from 129/sv mouse ES cells’ genomic DNA, mutated in exon22 encoding E988A, and cloned into the pEMC-neo targeting vector with a PGK-neo-resistance (neo-R) cassette flanked by FRTs. Targeted ES clones were screened by PCR first and confirmed via Southern blotting (HindIII and KpnI digestion with both 5′ and 3′ probes) (Fig. 1B and SI Appendix, Table S1). The successfully targeted ES clones were injected for germline transmission. The Parp1^+/AN chimeras with the neo-R cassette were bred with Rosa26a^FLIP/FLIP (Jax Strain No. 003946 in 129Sv) mice to generate the F1 Parp1^+/A mice. The presence of the NeoR in Parp1^AN allele interferes with the expression of Parp1, resulting in no or little Parp1 expression. The P-value of birth rates was calculated using χ² test. Genotypings were carried out by PCR with 230-bp product for Parp1⁺, 315 bp for Parp1^A, and440 bp for Parp1^AN in the triplex PCR, and 340 bp for only AN (Fig. 2 G and H and SI Appendix, Figs. S1 C and D and S2 C and D and Table S1). All animal work was carried out with the guidelines established by the Institutional Animal Care and Use Committee at Columbia University Medical Center.

Western Blot Antibodies.

Primary antibodies used in this study for western blot include: anti-PARP1 antibody (Cell Signaling Technology, 9542, 1:2,000), anti-PARP2 antibody (Active Motif, 39044, 1:2,000), anti-α-Tubulin antibody (Sigma, CP06, 1:1,000), anti-pan-ADP-ribose binding reagent (Sigma, MABE1016, 1:1,000), anti-β-actin (Sigma, A5441, 1:10,000), and anti-PAR antibody (R&D, 4335-MC-100, 1:1,000).

Mouse Hematopoietic Cell and Lymphocyte Analyses.

Mouse blood cell and lymphocyte development were analyzed by flow cytometry (36, 37). B lymphocyte development was conducted by staining with cocktails including FITC anti-mouse CD43 (Biolegend, 553270), PE goat anti-mouse IgM (Southern Biotech, 1020-09), PE-cyanine5 anti-Hu/Mo CD45R (B220) (eBioScience, 15-0452-83), and APC anti-mouse TER119 (Biolegend, 116212). T lymphocyte development was measured on thymocytes using cocktail including PE rat anti-mouse CD4 (BD Pharmingen, 557308), FITC anti-mouse CD8a (Biolegend, 100706), PE/Cy5 anti-mouse CD3e (eBioscience, 15-0031-83), and APC anti-mouse TCRβ (BD Pharmingen, 553174). Myeloid cells were measured in bone marrow cells and splenocytes using cocktail including FITC anti-mouse CD11b (BD Pharmingen, 553310), PE rat anti-mouse CD19 (BD Pharmingen, 557399), PE/Cy5 anti-mouse CD3e (eBioscience, 15-0031-83), and APC anti-mouse Ly6G/Ly6C(Gr-1) (Biolegend, 108412). Purified splenocyte CSR was stained with PE-cyanine5 anti-Hu/Mo CD45R (B220) and FITC rat anti-mouse IgG1(BD Pharmingen, 553443). All antibodies were diluted according to the manufacturer’s protocol. The flow cytometry data were collected on either an LSR II (BD), or on an Attune NxT (Invitrogen) flow cytometer. All flow cytometry data were analyzed using FlowJo V10.

Cell Line Derivation and Sensitivity Assay.

The Parp1^+/+, Parp1^−/−, Parp1^+/−, and F1 Parp1^+/A MEFs were derived from E13.5 to E14.5 embryos and functionally and genotyping verified (SI Appendix, Fig. S2 C and D). The Parp1^+/+, Parp1^−/−, Parp1^+/− and Parp1^+/A ES cells were derived at the HICCC transgenic mouse core facility and cultured on irradiated (30 Gy) fibroblast feeders as detailed before (38, 39). For the survival assay, ~1,000 cells/well were plated into a 96-well plate. Cells were irradiated by ionizing radiation (IR, 1, 2.5, or 5 Gy) 24 h after plating. Genotoxins were added 24 h after initial plating at the concentration indicated in the figures–camptothecin (CPT, Sigma, 208925) 50, and 100 nM; etoposide (Sigma, E1383) 50, 100, 250, 500, and 1,000 nM; and MMS (Sigma, 129925) 50, 100, and 200 μM. The CyQUANT™ Cell Proliferation Assay kit (Invitrogen, C7026) was used to measure cell numbers after 7 d. The survival curves were generated with GraphPad Prism V8.0 using the nonlinear regression linear quadratic cell death model, and P-value was calculated based on coefficient of killing.

Adult Male Mouse Epididymal Sperm Count and Sperm DNA Preparation.

Mice sperm counts and histology analyses of the adult testis (<8 wk) were carried out as detailed before (40). Testes were fixed in Bouin’s fixative (ICCA, 1120-16) for Periodic Acid Schiff staining. Cauda epididymis was minced in PBS and incubated at 32 °C for 20 min to let all sperms swim out. Sperm suspension was filtered through a 70-μm nylon cell strainer and counted under a microscope using a hematocytometer. Sperm mobility was determined as: grade a: immotile, b: nonprogressive, c: slow progressive, or d: rapid progressive.

Live-Cell Imaging.

Live-cell imaging and FRAP were conducted as previously described with minor revisions (10, 11). Plasmids encoding GFP-tagged WT or E988A PARP1, together with RFP-tagged XRCC1 (a generous gift from Dr. Li Lan at MGH), were transfected into PARP1 knockout U2OS cells (11) using Lipofectamine 2000 (Invitrogen, 11668019). Live-cell images were acquired via the NIS Element High Content Analysis software (Nikon Inc.) at different times after damage (inflicted by 405-nm laser without priming) and every 2.5 s after photobleaching (via a 488-nm laser) at 1 min after initial damage for FRAP. P-values of exchange curves were calculated based on extra sum-of-square F test using the dose–response model. All Dots and error bars represent means and SEs. All images’ analyses were carried out with Fiji ImageJ software.

HR and NHEJ Reporter Assay.

The DR-GFP, EJ5, and I-Scel plasmids were generous gifts from Dr. Jeremy Stark at the City of Hope, and the assays were carried out as previously detailed (28) with minor modifications. Briefly, the DR-GFP plasmid contains an upstream SceGFP cassette in which the GFP expression was interrupted by an I-Scel recognition site, and a downstream GFP template truncated at both termini. After I-Scel induced DSB, a successful HR repair using the downstream iGFP as template would convert the SceGFP cassette into an expressing GFP cassette. The EJ5 plasmid consists of a GFP cassette downstream of a puromycin ORF flanked by two I-Scel sites following a single promoter, in which the stop codon in the puromycin ORF prevents GFP expression. Successful NHEJ repair following I-Scel digestion restores GFP expression. Both the DR-GFP and EJ5 plasmids were linearized and electroporated (BioRad Gene Pulser Xcell™ system, 800 V, 10 μF, 300 Ω, 0.4-cm cuvettes) into immortalized Parp1^+/+, Parp1^−/−, F1 Parp1^+/A, and Xrcc4^−/− MEFs separately. After 6 d of puromycin selection and expansion, I-Scel plasmid was transiently transfected into the stable cell lines using lipofectamine 2000 following the manufacturer’s protocol. Frequency of GFP expressing cells was measured by flow cytometry at D3 after I-Scel transfection, and HR and NHEJ efficiency was calculated by normalizing GFP+ population to the efficiency of GFP expressing vector transfection control to elucidate the effect of transfection efficiency. All reporter assays were repeated for additional 2 times, and all flow cytometry data were analyzed using FlowJo V10.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

There are no data underlying this work.