Polynucleotide kinase–phosphatase (PNKP) is a DNA repair factor possessing both 5′‐kinase and 3′‐phosphatase activities to modify ends of a DNA break prior to ligation. Recently, decreased PNKP levels were identified as the cause of severe neuropathology present in the human microcephaly with seizures (MCSZ) syndrome. Utilizing novel murine Pnkp alleles that attenuate expression and a T424GfsX48 frame‐shift allele identified in MCSZ individuals, we determined how PNKP inactivation impacts neurogenesis. Mice with PNKP inactivation in neural progenitors manifest neurodevelopmental abnormalities and postnatal death. This severe phenotype involved defective base excision repair and non‐homologous end‐joining, pathways required for repair of both DNA single‐ and double‐strand breaks. Although mice homozygous for the T424GfsX48 allele were lethal embryonically, attenuated PNKP levels (akin to MCSZ) showed general neurodevelopmental defects, including microcephaly, indicating a critical developmental PNKP threshold. Directed postnatal neural inactivation of PNKP affected specific subpopulations including oligodendrocytes, indicating a broad requirement for genome maintenance, both during and after neurogenesis. These data illuminate the basis for selective neural vulnerability in DNA repair deficiency disease.
Polynucleotide kinase‐phosphatase (PNKP), a repair factor that modifies DNA breaks prior to ligation, is essential for normal brain development by preventing the accumulation of endogenous DNA breaks, which otherwise results in widespread cell death during neurogenesis.
PNKP is essential for neurogenesis by preventing DNA damage accumulation.
PNKP is required for efficient base‐excision repair (BER) of single‐strand breaks as well as non‐homologous end joining (NHEJ) repair of double‐strand breaks.
Attenuated PNKP levels generated using a hypomorphic Pnkp allele compromised brain development and generate a microcephaly with seizures (MCSZ)‐like neural phenotype.
Postnatal inactivation of PNKP in various neural compartments resulted in defective cellular homeostasis in mature cell types in the brain, including select neurons and oligodendrocytes.
Normal development and tissue homeostasis strictly depends upon the maintenance of genome integrity. This is key in all tissues, although the nervous system is particularly susceptible to genotoxic stress as revealed by many examples of neuropathology in human inherited DNA repair disorders (McKinnon, 2013; Madabhushi et al, 2014). While defects in each of the major DNA repair pathways impact multiple organ systems, disease‐causing mutations in the base excision repair/single‐strand break repair (BER/SSBR) pathway appear to uniquely affect the nervous system (Caldecott, 2008; McKinnon, 2009, 2013; Iyama & Wilson, 2013).
Base excision repair is critical for repair of frequent DNA damage events such as those arising from oxidative stress and involves multiple components that are assembled by XRCC1, a key scaffold factor in this pathway (Almeida & Sobol, 2007; Caldecott, 2008). Among these, poly(ADP‐ribose) polymerase (PARP) is an enzyme that activates signaling by multiple ADP‐ribosylation events (Caldecott, 2008; McKinnon, 2013). Other central components include apurinic/apyrimidinic endonuclease 1 (APE1) to initiate the repair of oxidative DNA lesions and polynucleotide kinase–phosphatase (PNKP), an enzyme critical for processing DNA ends prior to DNA ligation (Caldecott, 2008; Iyama & Wilson, 2013). Other end‐processing factors such as Aprataxin (APTX) and tyrosyl DNA phosphodiesterase 1 (TDP1) participate in the modification of specific DNA lesions, such as adenylation intermediates or trapped topoisomerase‐1 complexes (Ahel et al, 2006; Caldecott, 2008).
Among the BER machinery, PNKP, which consists of an N‐terminal forkhead‐associated (FHA) domain and a C‐terminal region that harbors a fused phosphatase and kinase region, hydrolyzes 3′‐phosphate groups and promotes phosphorylation of 5′‐OH termini (Pheiffer & Zimmerman, 1982; Karimi‐Busheri & Weinfeld, 1997; Jilani et al, 1999). Of these dual activities, the phosphatase activity appears to be the essential function of PNKP as the efficient repair of DNA lesions resulting from oxidative damage (e.g., DNA ends containing 3′‐phosphoglycolate) critically depends upon the 3′‐phosphatase function of PNKP (Breslin & Caldecott, 2009), via its interaction with XRCC1. Notably, PNKP has also been shown to associate with XRCC4, a factor involved in the non‐homologous end‐joining (NHEJ) pathway that repairs DNA double‐strand breaks (DSBs), implicating PNKP in both SSBR and DSBR (Chappell et al, 2002; Karimi‐Busheri et al, 2007; Segal‐Raz et al, 2011; Zolner et al, 2011). However, the physiologic role of PNKP and its relevance during either BER or double‐strand break repair (DSBR) is unclear.
Despite the essential nature of the BER pathway, there is clearly a difference in tissue requirements for individual components of this pathway. This likely reflects the nature of the damage and/or the threshold for genotoxic stress‐induced effects of different cells and tissues (Lee et al, 2012a; McKinnon, 2013). Recently, mutations in PNKP were identified as the cause of microcephaly with seizures (MCSZ), a syndrome characterized by profound neurodevelopmental microcephaly but without any obvious extra‐neurologic features (Shen et al, 2010; Poulton et al, 2013; Nakashima et al, 2014). Other disease‐causing mutations in DNA SSBR factors such as TDP1 or APTX result in pronounced neurodegeneration, but not microcephaly (Date et al, 2001; Takashima et al, 2002). An understanding of how individual components of this pathway differentially influence tissue homeostasis and how diseases of varied presentation and phenotype arise clearly requires physiologically relevant models. In this study, we have developed a series of unique genetic models for PNKP function that elucidate the requirements of this enzyme during neural development. In doing this, we uncovered important new insights into PNKP function in the nervous system, revealing a preeminent role in maintaining genome integrity during neurogenesis and in postmitotic neurons.
PNKP is essential for neurogenesis
To determine the physiologic requirements for PNKP, we generated multiple mutant Pnkp alleles to use as a tool to understand how mutation can lead to neurologic disease. We initially generated a conditional Pnkp allele by flanking exons 4 through 7 of mouse Pnkp with LoxP sites to enable cre‐mediated gene deletion (Fig 1A). This strategy generates an out‐of‐frame null allele, predicted to encode a truncated protein producing only an amino terminal peptide lacking any kinase or phosphatase domains. We subsequently generated mice with either embryo‐wide or tissue‐specific inactivation. We found that inactivation of Pnkp in the embryo using Sox2‐cre (which drives cre expression in the embryo proper and not in the placenta) resulted in early lethality. Mutant PnkpSox2‐cre embryos could not be recovered after embryonic day 9 (E9; not shown). Thus, like germ line deletion of other key components of BER such as XRCC1(Tebbs et al, 1999), PNKP is essential for early embryogenesis. In contrast, when Pnkp was deleted throughout neural development using Nestin‐cre (Fig 1B), we found that PnkpNes‐cre mice were born at Mendelian ratios, although these mutant mice did not survive past 5 days of age (Fig 1C). Thus, deletion of Pnkp even when restricted to the nervous system still resulted in lethality. These data highlight the essential role for PNKP during neural development and contrast the viability of mice after germ line inactivation of other DNA end‐processing factors such as APTX or TDP1 (Katyal et al, 2007). Even conditional neural deletion of XRCC1, while having a substantial impact on the brain, is viable for many months (Katyal et al, 2007; Lee et al, 2009).
To establish how PNKP loss impacts neural development, we first analyzed cortical development. The cortex is a laminar structure (Molyneaux et al, 2007), and perturbation of neural development often alters the arrangement of these six cortical layers. The PnkpNes‐cre cortex showed a marked reduction in size, although the ordered layering of the cortex still occurred (Fig 1D). However, there were reduced numbers of neurons, particularly in later‐born upper cortical layers (see Ctip2, Satb2 and Brn2 immunostaining; Fig 1E). We determined whether the reduced number of neurons was a result of cell death or decreased proliferation. During neural development, we observed abundant cell death from E13 onward using TUNEL analysis (Fig 2A). Cell death was widespread throughout the proliferating regions of the developing nervous system and was associated with DNA damage, indicated by increased γH2AX (Fig 2B). A reduction in proliferation was also found as evidenced by a markedly reduced level of BrdU incorporation (Fig 2C), although based on the high level of apoptosis, reduced proliferation is likely a secondary event. Conspicuously, PNKP loss is substantially more severe than inactivation of either LIG4 or XRCC1, indicating that this enzyme may be involved in the repair of a broader range of DNA lesions that either XRCC1 or LIG4 alone (Fig 2). To further assess PNKP function during corticogenesis, we also used Emx1‐cre, in which gene deletion occurs around 1 day earlier than Nes‐cre, and progenitors targeted by Emx1‐cre have a heightened sensitivity to DNA damage at this early stage (Lee et al, 2012a). Accordingly, in contrast to Nes‐cre, the cortices of PnkpEmx1‐cre mice were dramatically more affected, showing loss of all dorsal telencephalic progenitors via apoptosis, resulting in an almost complete absence of the cortex (Appendix Fig S1).
To ascertain the effector pathway responsible for cell loss, we generated PnkpNes‐cre mice that were also deficient for either p53 or ATM (Appendix Fig S2). P53 is an essential effector after DNA DSBs in all immature cells in the nervous system, while ATM is primarily required for DNA damage‐induced apoptosis in immature postmitotic neurons, rather than replication‐associated damage (Lee et al, 2001; Orii et al, 2006). We found that apoptosis in (Pnkp;Atm)Nes‐cre was identical to PnkpNes‐cre, while p53 inactivation abrogated apoptosis (Fig 3A). Consequently, (Pnkp;p53)Nes‐cre mice were born with a substantially more developed cortex compared with PnkpNes‐cre mice (Appendix Fig S2). This is also apparent by recovery of neuronal numbers and a more ordered cortical lamination as assessed by Satb2 and Ctip2 localization in (Pnkp;p53)Nes‐cre mice (Fig 3B). Thus, PNKP loss activates a p53‐dependent DNA damage‐induced apoptotic pathway.
Cerebellar neurons including the granule neurons in the EGL and particularly the interneurons (comprising the stellate, basket and golgi cells) are susceptible to multiple types of DNA damage. For example, interneurons fail to undergo postnatal expansion after XRCC1 loss due to a p53‐dependent cell cycle arrest, while inactivation of essential homologous recombination (HR) factors or telo‐mere uncapping results in p53‐dependent apoptosis (Lee et al, 2009, 2014). We found that PNKP inactivation also resulted in loss of this cerebellar neuronal population in an ATM‐independent, but (partially) p53‐dependent manner, with an approximately 40% recovery of interneurons (Fig 3C). Interneuron loss involved elevated γH2AX and apoptosis, indicated by increased TUNEL‐positive cells in the cerebellar white matter (Fig 3A). The partial interneuron rescue in the (Pnkp;p53)Nes‐cre cerebellum compared to XRCC1 inactivation, where associated p53 loss resulted in full interneuron recovery, likely reflects the requirement of PNKP for the repair of a broader range of DNA lesions (Lee et al, 2009, 2014) (Fig 1, and see below).
PNKP is critical for both BER and NHEJ
Polynucleotide kinase–phosphatase is a dual‐function kinase/phosphatase, which processes damaged DNA ends to enable ligation (Weinfeld et al, 2011). In particular, the 3′‐phosphatase activity is the critical functionality for SSBR after oxidative DNA damage (Breslin & Caldecott, 2009; Weinfeld et al, 2011). Polynucleotide kinase–phosphatase has also been linked to the repair of DSB lesions (Chappell et al, 2002; Karimi‐Busheri et al, 2007; Segal‐Raz et al, 2011; Zolner et al, 2011), although the direct importance of PNKP in DSBR is less firmly established than for BER. We reasoned that the increased impact of PNKP loss on neurogenesis (e.g., compared to XRCC1) reflects deficits in both BER and DSBR. To determine the importance of PNKP for repair of different DNA lesions, we established PnkpNes‐cre primary cortical astrocytes and those derived from other mutant mice defective in BER or NHEJ. Using the comet assay, we found a significant deficiency in the repair of DNA damage in PnkpNes‐cre cells after treatment with ionizing radiation (IR), hydrogen peroxide (H2O2) or camptothecin (CPT) in non‐replicating astrocytes, as shown by increased DNA strand breaks after recovery from genotoxin treatment (Fig 4A). Thus, loss of PNKP compromises DNA repair after a variety of genotoxins.
Because these cells were non‐replicating, H2O2 and CPT produce DNA SSBs, while IR produces both SSBs and DSBs. To establish whether the DNA repair deficiency in the PNKP‐null cells also involved defective DSBR, we used a neutral comet assay. In this experiment, we used bleomycin, an agent that generates a higher relative frequency of DNA DSBs to SSBs. We found that inactivation of PNKP reduces DSBR to levels similar to those after inactivation of the NHEJ factor, Lig4 (Fig 4B). In contrast, disabling the BER pathway via loss of XRCC1 did not impact repair of DSBs. Thus, PNKP is important for repair of DNA DSBs. We further determined whether PNKP loss involved defective NHEJ by inhibiting DNA‐PKcs in control and PnkpNes‐cre cells (Appendix Fig S3). DNA repair was reduced after inhibition of DNA‐PKcs using NU7026 in WT primary astrocytes (a repair defect comparable to DNA‐PKcs−/− astrocytes), while similar treatment of PNKP‐deficient cells showed no additional reduction in DNA repair capacity (Appendix Fig S3), suggesting the loss of PNKP was epistatic to defective NHEJ.
This finding was reinforced by γH2AX recovery assays to monitor DSBR, and showed that PnkpNes‐cre cells had a similar defect in DNA repair to LIG4 loss when compared to controls or those with XRCC1 inactivation (Fig 4C). We also observed widespread chromosomal lesions after IR as indicated by micronuclei and chromosome bridges in a high percentage of PNKP‐null cells compared to those with either BER or NHEJ deficiency (Fig 4D). Collectively, these data indicate that loss of PNKP impacts both BER and NHEJ.
PNKP is not required for homologous recombination
Given the DNA DSBR deficiency of PNKP‐null cells, we also determined whether PNKP was required for DSBR via HR. To do this, we exposed PNKP‐null cells to either mitomycin C (MMC) or cisplatin, both of which promote interstrand crosslinks, which requires HR for repair (Deans & West, 2011). We assessed RAD51 foci removal after DNA breaks as a measure of HR and included HR‐defective BRCA2‐null primary astrocytes as a control. We found that after MMC or cisplatin treatment, PNKP‐null cells were similar to controls in removal of RAD51 damage‐induced foci (Fig 5A; although basal γH2AX is higher in PNKP‐null cells). In contrast, BRCA2 null cells failed to assemble RAD51 foci and were also deficient in γH2AX resolution (Fig 5). Primary Atm−/− astrocytes were also included as an additional control as they had been reported to be defective in HR (Morrison et al, 2000), although recent data (Kass et al, 2013; Rass et al, 2013), and our current data (Fig 5), show that they are not HR deficient. A similar dispensability for PNKP during HR was also reported in glioma cells via analysis of sister chromatid exchange (Karimi‐Busheri et al, 2007). Thus, our findings indicate that PNKP is critical for chromosomal DNA repair via both BER and NHEJ, but not by HR.
The Pnkp MCSZ allele is lethal in the mouse germ line
The above data indicate that PNKP is an essential DNA repair factor utilized by different DNA repair pathways and that its loss impacts cellular genome stability in the face of endogenous DNA damage. Furthermore, the phenotype of the PnkpNes‐cre mouse highlights this enzyme as a key determinant for genome stability that is required for neural development. MCSZ results from hypomorphic mutation in PNKP and is notable as the symptomatic pathology is consistent with PNKP being involved in repair of DNA SSBs rather than DSBs (Shen et al, 2010; McKinnon, 2013; Poulton et al, 2013).
While the PNKP mutations in MCSZ target the kinase region, these mutations also lead to reduced levels of PNKP protein and overall enzyme activity, affecting 3′‐phosphatase activity (Shen et al, 2010; Reynolds et al, 2012). Therefore, we sought to determine how disruption of PNKP produces the characteristic neuropathology in MCSZ by generating a murine T424GfsX48 allele found in multiple MCSZ families (Shen et al, 2010; Poulton et al, 2013; Nakashima et al, 2014). This mutation generates a frame‐shift allele of PNKP resulting in the insertion of an additional 48 aa at residue 385 and a premature stop codon and truncated protein. Although we were able to identify the expected Pnkp mutant message in Pnkp+/T424GfsX48 heterozygous mice (levels of MCSZ transcript were reduced to around 20% of the WT allele), we could not recover any E9 (or later) homozygous PnkpT424GfsX48 mutant mice. We were also unable to rescue the lethality of homozygous mutant mice with concomitant loss of p53 or to recover homozygous mutants after outbreeding multiple generations to the highly robust NIMR mouse strain. Unlike humans, the early lethality of the MCSZ embryos suggests that the impact of reduced PNKP levels more substantially impacts early development in the mouse. Unfortunately, this early lethality precluded analysis of MCSZ mutation in the mouse nervous system. Thus, we conclude that reduced PNKP levels become critical in murine tissue, while the low PNKP levels found in humans support viability.
Reduced Pnkp expression causes neurodevelopmental defects
To generate a scenario mirroring the MCSZ syndrome in the mouse, we utilized an intermediate created during construction of the conditional Pnkp allele in which a Neo selection cassette was inserted into intron 4 (Fig 6A; this allele is termed PnkpNeo). Analysis of tissues from the PnkpNeo/Neo mice showed attenuation of PNKP levels (Fig 6B). Thus, similar to MCSZ in humans, PnkpNeo/Neo mice contained germ line attenuation of PNKP expression. In contrast to complete Pnkp deletion or homozygous expression of the MCSZ allele, PnkpNeo/Neo mice were viable and demonstrated neurodevelopmental deficits associated with reduced overall PNKP levels (Fig 6; Appendix Fig S4). While the reduced PNKP levels resulted in a smaller brain (see Fig 6B, arrows), analogous to that found in the human MCSZ syndrome, they were also associated with generalized growth defects. Thus, the relative impact of reduced Pnkp expression in mouse is more consequential than in humans. Nonetheless, these mice are valuable for understanding the physiologic requirement for PNKP during neural development.
Similar to PnkpNes‐cre mice, we found substantial apoptosis in the PnkpNeo/Neo brain from E13.5 that accounted for the smaller organ size (Fig 6C). Analysis of the developing E13.5 PnkpNeo/Neo neocortex showed an increase in DNA damage (γH2AX) and associated apoptosis (TUNEL) in mutant tissue compared to controls (Fig 6C). There was a concomitant reduction in proliferating cells (phospho‐histone H3) in both the apical and basal progenitor populations in the PnkpNeo/Neo tissue (Fig 6C, arrows), as proliferating cells are susceptible to DNA damage‐induced apoptosis.
We also determined DNA repair capacity in primary cells isolated from the hypomorphic PnkpNeo/Neo mice to compare with repair defects associated with reduced PNKP levels found in MCSZ (Shen et al, 2010; Reynolds et al, 2012). Because the PnkpNeo/Neo alleles are expressed in the germ line, we examined PnkpNeo/Neo mouse embryonic fibroblasts (MEFs) after various genotoxins (Fig 6D). We found reduced chromosomal DNA repair capacity in PnkpNeo/Neo MEFs compared to WT controls. Notably, we found some differential sensitivity to genotoxins, as the PnkpNeo/Neo MEFs were more defective in repair of IR or MMS damage compared with CPT; this was apparent in comparison with the Lig4−/− and Tdp1−/− mutants that were used as positive (repair‐deficient) controls, and probably relates to the relative requirements for PNKP in different end‐processing events (Fig 6D). We also used neutral comet assays to measure DNA DSBs after IR or bleomycin and found comparable DNA repair defects in PnkpNeo/Neo cells to those present in Lig4‐null cells, further illustrating defective DSBR after compromised PNKP levels (not shown). These data suggest that reduced PNKP levels in the PnkpNeo/Neo mice create a scenario much like that in MCSZ, where PNKP levels are sufficient for viability, but compromise genome integrity in the face of endogenous DNA lesions, leading to neural cell loss and resultant neuropathology.
PNKP maintains differentiated neurons
Little is known regarding the ongoing requirement for DNA repair factors in differentiated neural populations after the completion of neurogenesis (McKinnon, 2013). However, this is likely to be important for neural homeostasis, and to prevent progressive decline observed in many DNA repair deficiency syndromes. Therefore, we generated mice in which a tamoxifen (TX)‐inducible promoter was used for Pnkp inactivation in different neural compartments after neurogenesis. We initially examined the effect of postnatal Pnkp deletion using CreTM, an actin‐based promoter that is effective for broad neural gene inactivation (Hayashi & McMahon, 2002). We induced deletion at P21 using TX (a dose every 2 days, for a total of 3 doses) and examined tissue approximately 2 weeks later (Fig 7A). Compared to PnkpNes‐cre animals, TX treatment had a relatively modest effect toward the PnkpCreTM brain, with little discernable gross tissue effect (Fig 7A). This lack of an overt effect after PNKP loss is likely because cells were non‐replicative, thereby avoiding apoptosis from DNA damage during proliferation.
Inactivation of PNKP was confirmed using Western blot analysis, which showed reduced PNKP levels in TX‐treated PnkpCreTM brain tissue (Fig 7B). Under the induction conditions used, effective PNKP loss occurred throughout the cortex, while loss was more modest throughout the cerebellum. When cre induction began at earlier stages such as P7, the overall consequences of PNKP inactivation were more pronounced, in accord with increased neurogenesis at this earlier time (Appendix Fig S5). Nonetheless, deleterious effects of Pnkp inactivation toward multiple neural cell types were observed by P33 after TX administration at P21 (Fig 7C). For example, myelin basic protein (MBP), a central component for myelination by oligodendrocytes (Emery, 2010), was found to be markedly reduced throughout the PnkpCreTM brain, by both immunostaining and Western blot analysis (Fig 7B and C). Additionally, we found that the level of NeuN, an RNA splicing factor (Kim et al, 2009) also known as Fox‐3, which is expressed in mature neurons, was strongly attenuated (Fig 7B). To investigate whether the reduction in MBP is associated with oligodendrocyte loss, we used Sox10, which is a specific marker of oligodendrocytes in the central nervous system (Stolt et al, 2002). We found that numbers of Sox10‐positive cells were similar between control and the TX‐induced PnkpCreTM cortex, although there was a reduction in cell numbers in the dentate gyrus, a structure that maintains neurogenesis until ~P36 (Appendix Fig S6). Similarly, the observed NeuN decrease was not associated with loss of neurons as evident by similar DAPI staining between control and mutant tissues (Fig 7C, white arrowheads). These data indicate a broad impact of PNKP inactivation throughout the postmitotic brain, although not involving overt cell loss as found during neurogenesis. In contrast, Pnkp deletion after TX treatment between P7 and P14, when neurogenesis is still occurring in the dentate gyrus and cerebellum, showed a more pronounced effect involving neurogenesis‐associated cell loss (Appendix Fig S7). Again, we also found a clear effect of PNKP loss toward oligodendrocytes, with a reduction in MBP by P20 after induction of Pnkp inactivation at P7 (Appendix Fig S6). Thus, the genome maintenance functions of PNKP are critical for the homeostasis and survival of postnatal neural populations.
PNKP is an essential factor for glial homeostasis
Based on the oligodendrocyte sensitivity data above, and because the relative in vivo requirements for the DDR in glial cells are unclear, we used an inducible GFAP‐cre promoter (Chow et al, 2008) to generate PnkpGfap‐creTM mice to further assess PNKP in selective neural compartments. GFAP‐creTM is a transgene that directs cre expression throughout glia in the adult brain, but with only sporadic expression in the radial glia progenitors in the subventricular zone progenitors of the lateral ventricle, thereby mostly restricting gene deletion to differentiated glia (Chow et al, 2008).
We found that even with early deletion at P1, the loss of PNKP in PnkpGfap‐creTM produced a less severe phenotype than PnkpCreTM (Appendix Fig S5), with overall brain and body size being similar to controls, although a modest reduction in cortical size was found in PnkpGfap‐creTM brains (Fig 8A). We again observed an oligodendrocyte defect in PnkpGfap‐creTM mice as determined by decreased immunostaining for MBP (Fig 8B). We also observed altered NeuN immunostaining patterns in cortical neurons, which given the lack of neuronal expression of the GFAP‐creTM transgene, likely reflect an indirect effect from loss of myelination (Fig 8B). Other markers of oligodendrocytes such as MBP, 2′,3′‐cyclic‐nucleotide 3′‐phosphodiesterase (CNPase) and Olig2 are also affected by PNKP inactivation (Fig 8C). These effects did not involve oligodendrocyte cell loss, as Sox10 immunopositive cell numbers were similar between PnkpGfap‐creTM and control brains (Fig 8D). In contrast to oligodendrocytes, loss of astrocytes was found in PnkpGfap‐creTM brain, reflecting ongoing proliferation in these cells (Appendix Fig S8). Collectively, our data indicate that attenuated PNKP levels cause pronounced neuropathology, which is characterized by retarded neural development and microcephaly resulting from DNA damage during neurogenesis. PNKP also fulfills important genome maintenance functions after the completion of neurogenesis to preserve neural homeostasis, as loss of PNKP strongly impacts oligodendrocytes, which may account for white matter abnormalities that are apparent as disease progression occurs in MCSZ (Poulton et al, 2013).
The MCSZ phenotype results from germ line PNKP mutations and extensively affects the nervous system. Given the ubiquitous tissue distribution of the BER pathway, this essential neural function of PNKP implicates increased levels or specific types of endogenous DNA lesions (e.g., those arising from oxidative damage), possibly coupled to a lower threshold of cellular tolerance in this tissue. To determine how PNKP maintains genome integrity in the setting of the nervous system, we adopted a multipronged approach to understand the functional relevance of this enzyme to neural homeostasis and disease. Here we provide a comprehensive analysis of the in vivo physiologic role of PNKP during development. Our data showed that the MCSZ phenotype is attributable to endogenous DNA damage and reduced repair capacity because of decreased PNKP levels, resulting in neural cell death and microcephaly. Furthermore, we also show that there is an ongoing role for PNKP function after the completion of neurogenesis, which is critical in the glial cell population. Thus, our data demonstrate that reduced PNKP function in the face of endogenous DNA damage widely impacts the nervous system resulting in a phenotype reflective of MCSZ.
In other neurologic syndromes, loss of SSBR factors leads to neurodegeneration without extra‐neurologic consequences, while deficiency in DNA DSBR results in neurodegeneration or microcephaly associated with broad defects throughout the body (McKinnon, 2009, 2013). While PNKP is a key factor in BER via its interaction with the SSBR factor XRCC1, it has also been shown to associate with XRCC4, a factor involved in the NHEJ pathway that repairs DNA DSBs, thereby implicating PNKP in both SSBR and DSBR (Koch et al, 2004; Caldecott, 2008). Our data clearly show a role for PNKP in both BER and NHEJ (but not HR) consistent with prior reports involving cell reconstitution studies and RNA interference experiments (Chappell et al, 2002; Koch et al, 2004; Karimi‐Busheri et al, 2007). However, as the repair defects found in Pnkp−/− cells after IR are similar to those seen after XRCC1 or LIG4 loss rather than being substantially greater, it is likely that PNKP may participate in a fraction of the breaks during NHEJ repair of IR damage, rather than all breaks as would be expected for LIG4. Given the importance for PNKP during both BER and NHEJ, one would predict deficiency of this factor would impact outside the nervous system similar to human syndromes characterized by DSBR defects such as Nijmegen breakage syndrome or Lig4 syndrome (Girard et al, 2004; Ben‐Omran et al, 2005; O'Driscoll & Jeggo, 2006). This may indicate that in MCSZ, specific types of abundant neural DNA lesions (arising from high oxygen consumption?) depend on normal PNKP levels, compared to other general DNA lesions.
In contrast to the human disease, the observation that our engineered mouse MCSZ mutation is incompatible with embryo viability suggests that different thresholds exist for PNKP function between human and mouse. However, phenotypic variation exists in humans harboring the T424GfsX48 allele; although the MCSZ phenotype is microcephaly (Shen et al, 2010), this allele was recently linked to neurodegeneration and cerebellar ataxia in another family with two affected siblings (Poulton et al, 2013). Thus, these phenotypic differences in human MCSZ may suggest genetic modifiers or background can affect the outcome of PNKP inactivation. Possibly, the mouse strains used in our study predispose to higher sensitivity to PNKP loss?
We also identified an ongoing requirement for PNKP function in mature neuronal and glial cells, by inducing gene deletion after the cessation of neurogenesis. Inactivation of PNKP in non‐cycling SOX10‐positive oligodendrocytes led to loss of myelination, indicating endogenous DNA damage impacts mature glial function. As oligodendrocytes are sensitive to oxidative stress (Dewar et al, 2003; Jana & Pahan, 2007; French et al, 2009), this may indicate a key role for PNKP for DNA repair of oxidative DNA damage. PNKP requirement in glia is also notable in light of a report indicating that glia do not activate a DNA damage response (Schneider et al, 2012). We have previously shown that glia do activate a normal DDR if their genome integrity is compromised (Lee et al, 2012a, 2014). DNA damage‐induced disruption of normal oligodendrocyte function may well underpin white matter defects typical of MCSZ (Shen et al, 2010; Poulton et al, 2013). In fact, this sensitivity of oligodendrocytes raises the possibility that genomic damage and genotoxins may contribute to demyelination defects in more common neurologic disease such as multiple sclerosis (MS) (Lee et al, 2012b). This is particularly relevant as oligodendrocyte turnover in humans is very limited, while myelination changes are dynamic and contribute to neural plasticity (Yeung et al, 2014). Importantly, as the initiating lesions that underpin MS are relatively unknown (Lassmann et al, 2012; Friese et al, 2014), links between genome damage and white matter/myelin defects warrant closer inspection.
In the case of neurons, we found a marked attenuation of NeuN levels (an RNA splicing factor expressed in mature neurons), indicating neural homeostasis is perturbed. However, given that we found an effect on NeuN‐positive neurons after Pnkp deletion via Gfap‐creTM, glial defects may contribute more broadly in the mature brain to ongoing neuropathology. The oligodendrocyte/MBP sensitivity, and altered NeuN expression, suggests that disease progression may also involve disruption of transcription as a consequence of DNA damage. Transcription‐associated R‐loops are potential sources of DNA breaks that can compromise genome stability, highlighting the need for DNA repair in this setting (Mischo et al, 2011; Aguilera & Garcia‐Muse, 2012; Gaillard et al, 2013; Bhatia et al, 2014). Transcriptional disruption has been suggested to underpin the neurodegenerative phenotype associated with AOA2, SCAN1 and other neurologic diseases resulting from genomic damage (Suraweera et al, 2009; King et al, 2013; Richard et al, 2013; Yuce & We st, 2013; Gomez‐Herreros et al, 2014).
Thus, PNKP is an essential DNA repair factor that participates in both DSBR and SSBR, and is necessary to prevent severe neurologic decline in humans. Our data indicate that genome stability afforded by PNKP is critical both during neurogenesis and in the mature brain. These findings point to the ongoing need for multiple genome stability mechanisms at all stages of the nervous system, underscoring the critical links between genome stability and the prevention of neurologic disease.
Materials and Methods
Pnkp mutant mice
Two individual targeting vectors were constructed to generate a conditional Pnkp allele or the MCSZ mutant allele. The Pnkp conditional strategy resulted in exons 4–7 flanked by LoxP sites. To achieve this, we used recombineering to introduce a single LoxP site between exons 7 and 8 and a neomycin selection cassette flanked by Frt sites with a single LoxP site was inserted between exons 3 and 4. The targeting construct was electroporated into W9.5 embryonic stem (ES) cells, and after selection with G418, homologous recombinants were screened by Southern blotting using both 3′ and 5′ probes outside the targeted region. Positive clones were blastocyst‐injected to generate chimeric mice. The male chimeras were bred with C57BL/6 females to establish mice carrying the targeted Pnkp allele. The initial mouse line still contained the neomycin cassette inserted in intron between exons 3 and 4, PnkpNeo. Because the homozygous PnkpNeo/Neo mice had a significant phenotype, they were maintained as an independent line (this mutation was germ line; therefore, these mice were not crossed with cre lines). Further breeding with Flp recombinase mice (JAX #003800: B6; SJL‐TG(ACTFLPE)9205 DYM/J) removed the neomycin cassette resulting in the conditional PnkploxP line.
Further breeding with cre mice under the control of the Nestin promoter (JAX #003771: B6.Cg‐Tg(Nes‐cre)1Kln/J) resulted in deletion throughout the nervous system (http://cre.jax.org/Nes/Nes-CreNano.html), while breeding with the Emx1‐cre mice (JAX #005628; B6.129S2‐Emx1tm1(cre)Krj/J) deleted Pnkp in Emx1‐expressing neural progenitors. Sox2‐cre mice (JAX #004783; Tg(Sox2‐cre)1Amc) were used to delete Pnkp in the embryo proper (avoiding placental inactivation). CreTM (JAX #004682; B6.Cg‐Tg (CAG‐cre/Esr1)5Amc/J), Rosa26CreTM (JAX 008463: B6. 129‐Gt (ROSA) 26Sortm1(cre/ERT2)Tyj/J) and GFAP‐CreTMER2 (Chow et al, 2008) provided tamoxifen‐inducible cre expression driven from the actin or GFAP promoter, respectively, and were used to obtain primary cell cultures.
The MCSZ mutant mice were created using recombineering; an initial targeting introduced the 17‐bp mutation into exon 13 leaving a LoxP site in the preceding intron (between exons 12 and 13). The neomycin–thymidine kinase dual selection cassette flanked by Frt sites with a single LoxP was inserted between exons 3 and 4. A second round of selection in ES culture following transfection with Flp recombinase and using FIAU resulted in a clone which was subsequently injected into blastocysts and the resulting chimeric males bred with C57BL/6 females to establish mice carrying the mutant Pnkp allele.
Other mice used were p53LoxP, AtmLoxP, Lig4LoxP and Xrcc1LoxP which have been described previously (Herzog et al, 1998; Lee et al, 2009; Shull et al, 2009). The Atm conditional allele was engineered to have exon 58 flanked by LoxP sites (Lee et al, 2012a). All animal experiments were carried out according to NIH regulations and were approved by the SJCRH Animal Care and Use Committee.
Tamoxifen‐induced cre expression
Tamoxifen administration was done via intraperitoneal injection using tamoxifen (Sigma) dissolved in corn oil (Sigma) at 10 mg/ml. Tamoxifen injections were done with PnkpRosa‐CreTM, PnkpCreTM, PnkpGfap‐CreTM and Xrcc1Rosa‐CreTM mice using a dose of 1 mg/10 g mouse body weight as per the indicated dosing schedules.
Histology and immunodetection
Mice underwent transcardial perfusion with 4% (w/v) buffered paraformaldehyde (PFA), or embryos were drop‐fixed in 4% PFA. Brains or embryos were cryoprotected in buffered 25% sucrose (w/v) solution. After embedding in Tissue‐Tek OCT compound (Sakura), samples were cryosectioned sagittally at 10 μm using a CM3050S cryostat (Leica). Immunohistochemical analysis of sectioned samples was first subjected to antigen retrieval according to the manufacturer's directions (HistoVT One, Nacalai Tesque). For all immunohistochemistry and immunocytochemistry, the following antibodies (in alphabetical order) were used: 53BP1 (rabbit, 1:500, Bethyl Labs, cat# A300‐272A), Brn2 (rabbit, 1:200, Gene‐Tex, cat# GTX114650), CNPase (mouse, 1:500, Sigma, cat# C5922), Ctip2 (rat, 1:100, Abcam, cat# ab18465), Cux1 (CDP, rabbit 1:100 Santa Cruz, cat# SC13024), GFAP (mouse, 1:500, Sigma, cat# G3893), γH2AX (Ser139, rabbit, 1:500, Cell Signaling, cat# 2577, and mouse, 1:500, Millipore, cat# 05‐636), phospho‐histone H3‐Ser10 (mouse, 1:1,000, Cell Signaling, cat# 9706S), MBP (rabbit, 1:500, Abcam, cat# ab40390), NeuN (mouse, 1:500, Chemicon, cat# MAB377), Olig2 (rabbit, 1:500, Millipore, cat# AB9610), Pax2 (rabbit, 1:500, Zymed, cat# 71‐6000) PCNA (clone PC10, mouse, 1:500, Santa Cruz, cat# SC‐56), RAD51 (rabbit, 1/500, Santa Cruz, cat# sc8349), Satb2 (mouse, 1:50, Abcam, cat# 51502), Sox2 (rabbit, 1:500, Millipore, cat# AB5603), Sox10 (rabbit, 1:250, Abcam, cat# ab155279), Tbr1 (rabbit, 1:100, Abcam, cat# ab31940) and Tuj1/β‐tubulin III (mouse, 1:1,000, Covance, cat# MMS‐435P). For colorimetric visualization of positive signals, sections were incubated with primary antibodies overnight at room temperature, and endogenous peroxidase was quenched using 0.6% (v/v) H2O2 in methanol and blocked with goat serum [5% (v/v) goat serum, 1% (w/v) BSA in PBS‐T] at room temperature for 1 h. Slides were washed three times with PBS followed by a 1‐h incubation with biotinylated secondary antibody and avidin–biotin complex (Vectastain Elite kit, Vector Labs). Immunoreactivity was visualized with the VIP substrate kit (Vector Labs) using the manufacturer's protocol. Sections were counterstained with 0.1% (w/v) methyl green, dehydrated and mounted in DPX (Fluka). For fluorescent detection of immunoreactivity, FITC‐ or Cy3‐conjugated secondary antibodies (Jackson Immunologicals) were used and counterstained with 4′,6‐diamidino‐2‐phenylindole (DAPI; Vector Laboratories).
For BrdU incorporation assays, pregnant mice at E13.5 or E15.5 were injected (60 μg per 1 g of body weight), and 2 h after injection, embryos were drop‐fixed in 4% PFA and processed for cryosectioning and anti‐BrdU (rat, 1:200, Abcam, cat# ab6326‐250) immunostaining was assessed. Apoptosis was measured by TUNEL assay using Apoptag (Millipore) according to the manufacturer's protocol. Following TUNEL staining, sections were counterstained with propidium iodide (PI) in mounting medium (Vector Laboratories). Nissl staining was performed with 1% (w/v) thionin using standard protocol; Hematoxylin/eosin staining was also performed using standard procedures.
Isolation of primary cells and immunofluorescence
Primary astrocytes were prepared from P1 to P5 mouse brains as described previously (Katyal et al, 2007). Cortices were dissected free of meninges and then dissociated by passage through a 5‐ml pipette, and cells were resuspended in Dulbecco's modified Eagle's medium and Ham's nutrient mixture F‐12 (1:1 DMEM/F12, Gibco‐BRL) supplemented with 10% fetal bovine serum (v/v), 1× glutamax, 100 U/ml penicillin, 100 mg/ml streptomycin and 20 ng/ml epidermal growth factor (EGF; Millipore). Primary astrocytes were established and maintained in Primaria T‐25 tissue culture flasks (Falcon) at 37°C in a humidified CO2‐regulated (5%) incubator.
Primary MEFs were prepared from E13.5 embryonic mesenchyme. Tissue was minced using dissection scissors, trypsinized and resuspended in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (v/v), 1× glutamax, 100 U/ml penicillin, 100 μg/ml streptomycin and β‐mercaptoethanol and established in T‐25 tissue culture flasks (Falcon) at 37°C in a humidified CO2‐regulated (5%) incubator.
For fluorescent labeling of quiescent cortical astrocytes or MEFs, cells were grown to confluence on glass coverslips, fixed with 4% PFA in PBS for 10 min and permeabilized for 5 min in 0.5% Triton X‐100/PBS. Cells were immunostained as described above, followed by Alexa 488/555‐conjugated secondary antibodies and counterstained with DAPI.
Alkaline comet assay
Cells were treated with either 150 μM H2O2 (Thermo Fisher) on ice for 5 min, methyl methanesulfonate (MMS, Sigma) for 10 min at 0.2 mg/ml at 37°C, 14 μM camptothecin (CPT, Calbiochem) at 37°C for 30 or 60 min, bleomycin (Bedford Laboratories) at 40 μg/ml for 30 at 37°C, or γ–irradiation (10 Gy for alkaline comet assay or 20 Gy for neutral comet assay using 137Cs). After this step, all subsequent steps were performed in the dark, and comet assays were done as previously described (Katyal et al, 2007, 2014). Briefly, trypsinized cells were mixed with equal volume of 1.2% low‐melting‐point agarose (Invitrogen) and plated onto frosted glass slides (Fisher) pre‐coated with 0.6% agarose. Slides were treated with lysis buffer (2.5 M NaCl, 10 mM Tris–HCl, 100 mM EDTA (pH 8.0), 1% Triton X‐100, 3% DMSO, pH 10) for 1.25 h at 4°C and placed in pre‐chilled alkaline electrophoresis buffer (50 mM NaOH, 1 mM EDTA and 1% DMSO) for 45 min. Electrophoresis was done at 95 mA for 25 min and neutralized in 0.4 M Tris–HCl (pH 7.5). Comets were stained using SYBR Green (1:10,000 in 0.4 M Tris–HCl, pH 7.5) for 10 min. At least 100 comets/experiment were counted and measured using the Comet Assay IV system (Perceptive Instruments) together with an Axioskop2 plus microscope (Carl Zeiss) at ×200 magnification.
Neutral comet assay
Neutral comet assays were done as above, but using a sodium acetate‐based electrophoresis buffer (300 mM NaOAc, 100 mM Tris–HCl, pH 8.5); electrophoresis was performed with at 190 mA for 50 min followed by washing in neutralization buffer (0.4 M Tris–HCl pH 7.5).
Mouse tissues were harvested and extracted by RIPA buffer (0.1 M Tris–HCl (pH 7.5), 0.5 M NaCl, 0.05 M EDTA, 1% Triton X‐100, 1% SDS, 1% sodium deoxycholate) with protease inhibitors (1 mM NaF, 1 mM Na3VO4, protease inhibitor cocktail (Roche)). Proteins were quantified using Bradford standard method (Bio‐Rad). Protein samples were electrophoresed with 4–12% Bis‐Tris SDS–polyacrylamide gel (Invitrogen) and transferred to nitrocellulose membrane (Bio‐Rad). Membranes were blocked by 5% skim milk/TBS‐T at 37°C for 1 h. Primary antibody was incubated at room temperature for 4–24 h. Antibodies used for Western blotting were as follows: β‐actin (goat, 1:1,000, Santa Cruz, cat# sc1616), MBP (rabbit, 1:2,000, Abcam, cat# ab40390), NBS1 (rabbit, 1:1,000, Cell Signaling, cat# 3002) and PNKP (rabbit, 1:1,000, Novus, cat# NBP1‐87257). Secondary antibody was reacted at room temperature for 1 h. Membranes were developed using Amersham ECL Western blotting detection reagent (GE Healthcare Life Sciences). Ponceau S staining was used as a loading control.
Magnetic resonance imaging (MRI)
PnkpNes‐Cre, PnkpNeo/Neo and control mice (age at P2 and 8 months, respectively, sex matched) were used for MRI analysis. All scans were performed with 7 Tesla Bruker Clinscan animal magnetic resonance imaging scanner (Bruker Biospin MRI GmbH).
In situ cell counts
To determine γH2AX, BrdU, phospho‐H3 or TUNEL‐positive cell numbers in E13.5 and E15.5 embryonic cortex, immunostained sections were subjected to quantitative analysis. Three embryos for each genotype were analyzed. Immunopositive signals were measured within 200‐μm fields from at least three representative sections of the neocortex for each individual embryo.
Unpaired Student's t‐tests were used for the determination of P‐values. Values represented the mean and standard error of at least three independent experiments.
MS and PJM conceived and planned all project experiments and produced the final version of the manuscript. MS and LCD performed all experiments. HRR generated the Pnkp mutant mouse models and helped with experiments.
Conflict of interest
The authors declare that they have no conflict of interest.
We thank Dr. Suzanne Baker for providing the GFAP‐creTM mice and Susanna Downing for helpful comments and advice. We also thank Dr. Jingfeng Zhao and Yang Li for genotyping and technical help. We thank the Animal Resource Center, the Small Animal Imaging Center and the Transgenic Core Unit for support with mouse work. PJM is supported by the NIH (NS‐37956, CA‐21765), the CCSG (P30 CA21765) and the American Lebanese and Syrian Associated Charities of St. Jude Children's Research Hospital. MS is also supported by the Uehara Memorial Foundation of Life Sciences.
- © 2015 The Authors