The ZFHX3 (ATBF1) transcription factor induces PDGFRB, which activates ATM in the cytoplasm to protect cerebellar neurons from oxidative stress

Ataxia telangiectasia (A-T) is a neurodegenerative, inherited disease caused by mutations in the ATM gene (Savitsky et al., 1995), which encodes a serine-threonine kinase that plays a central role in the cellular response to DNA double-strand breaks (DSBs) (Khanna and Jackson, 2001; Shiloh, 2003). The name A-T represents the clinical symptoms, which include degeneration of the cerebellum causing ‘ataxia’ and microcapillary aneurysm giving rise to ‘telangiectasia’.

Abnormal elevation of alpha-fetoprotein (AFP) is one of the diagnostic markers of A-T patients. The AFP and albumin genes are present in tandem on human chromosome 4 (Kao et al., 1982; Urano et al., 1984). AFP is normally expressed exclusively in embryonic liver and completely suppressed in normal adult liver, but it is abnormally elevated in the serum of adult patients with A-T. The mechanism of gene suppression of AFP must be dysfunctional in A-T, and this mechanism might be involved in other systemic problems of the disease.

We focus here on the functional relationship between the symptoms of A-T and the transcription regulatory factor ZFHX3 (ATBF1). This factor was discovered as a DNA-binding protein that binds the AT (adenine and thymine)-rich element of the AFP gene to suppress its expression and was named ATBF1 (AT-motif binding factor 1) (Morinaga et al., 1991); subsequently, the full-length transcript ATBF1-A was found to be expressed in a neuronal differentiation-dependent manner (Miura et al., 1995; Jung et al., 2005). More recently, the Human Genome Organisation named the transcription factor as ZFHX3 (zinc finger homeobox 3). The binding of ZFHX3 to the AT-rich element of the AFP gene suppresses expression by interfering with the binding of activators (Morinaga et al., 1991; Yasuda et al., 1994). Interplay between p53 and TGF-β effectors also plays an important role in the suppression of the AFP gene (Wilkinson et al., 2008). Furthermore, p53 binds to and cooperates with ZFHX3 to activate the p21^Waf1/Cip1 promoter to trigger cell cycle arrest (Miura et al., 2004).

In addition to being elevated in adults with A-T, AFP is expressed in hepatocellular carcinoma (HCC) cells and in specific gastric cancer cells, called AFP-producing gastric cancer (AFP-GC) cells, in adults, and this has been linked with ZFHX3 deficiency. ZFHX3 mRNA expression in HCCs was significantly reduced in cancer cells compared with the corresponding surrounding liver tissues (Kim et al., 2008). In addition, AFP-GC cells showed reduced or absent ZFHX3 expression, and an extremely malignant character with a high frequency of metastasis, which might be related to altered function of adhesive molecules (Kataoka et al., 2001; Cho et al., 2007).

The most serious issue in A-T is the early onset of neurodegeneration in the cerebellum. DNA DSBs play an important role in inducing neurodegeneration in A-T. However, paradoxical clinical cases have been reported with mild symptoms, as defined by clinical examination and a quantitative A-T neurological index. Surprisingly, no ATM was detected in such patients’ cells, and sequence analysis revealed that they were homozygous for a truncating ATM mutation that is expected to lead to the classical, severe neurological presentation. Moreover, the cellular phenotype of these patients was indistinguishable from that of classical A-T: all the tested parameters of the DNA DSBs response were severely defective, as in typical A-T. This analysis showed that the severity of the neurological component of A-T was determined not only by ATM mutations but also by other influences yet to be found (Alterman et al., 2007). Researchers therefore started to search for mechanisms other then DNA DSBs to improve understanding of A-T neurodegeneration.

ZFHX3 was found to be one of several hundred substrates of ATM (Matsuoka et al., 2007). The functional meaning of its phosphorylation could be similar to the phosphorylation of p53 by ATM, because both ZFHX3 and p53 share the same consensus sequence as a target of ATM and function as a member of transcription factors in the nucleus. Phosphorylation stabilizes p53 by reducing the interaction with its negative regulator, the oncoprotein MDM2 (Banin et al., 1998; Shieh et al., 1997). We estimated that ATM might also stabilize and activate ZFHX3. We are interested in the clinical symptoms of A-T patients, which could be partly explained by the abnormal function of ZFHX3. Recently, variants in ZFHX3 have been reported to be associated with genetic susceptibility to Kawasaki disease (also known as mucocutaneous lymph-node syndrome), which induces high risk of coronary aneurysmal dilatation (Burgner et al., 2009). The report suggested that ZFHX3 played a role in protection against the formation of aneurysms.

The relationship between ZFHX3 and various symptoms of A-T suggests an important role of ZFH3 in the pathogenesis of A-T, and we therefore sought to understand its transcription and identify its target genes. We describe here a transcriptional regulatory mechanism that operates ZFHX3 gene expression in response to retinoic acid via ATM activation, and report that ZFHX3 regulates expression of adhesion molecules and platelet-derived growth factor receptor β (PDGFRB). Furthermore, to clarify the reason why A-T shows neurodegeneration specifically in the cerebellum, we investigated a distinct expression of PDGFRB in the cerebellum and analyzed the biological meaning of this expression. It has been known that the activation of p53 by oxidative stress involves PDGFRB-mediated ATM kinase activation (Chen, K. et al., 2003). Oxidative stress induces activation of ATM in the cytoplasm (Alexander et al., 2010). Collectively, these observations suggested an important role of ATM in the cytoplasm in response to oxidative stress in neurons independent from a role in repairing DNA DSBs in the nucleus (Dupre et al., 2006). We investigated a newly identified signal pathway that activates ATM in the cytoplasm for the protection of neurons against excitotoxicity. The mechanism should be responsible for the survival of neurons to protect organelles from oxidative stress in the cytoplasm.

Retinoic acid (RA) strongly affects the expression of Hox homeotic genes. Hox gene expression by neuronal stem cells in the hindbrain and branchial region of the head in the mouse embryo is particularly sensitive to RA (Marshall et al., 1996). The induction of neuronal differentiation by RA has been established using cell lines including human and mouse embryonal carcinoma cells (Andrews et al., 1984; Andrews, 1984). The regulatory mechanism of gene expression with RA treatment is based on nuclear retinoic acid receptors (RARs or RXRs) that recognize particular sequence motifs within target genes, called retinoic acid response elements (RAREs) (Marshall et al., 1996). ZFHX3 is a nonclustered homeotic factor expressed in post-mitotic neurons in the embryonic brain (Jung et al., 2005). Importantly, the ZFHX3 promoter is known to lack RAREs, but ZFHX3 expression is induced by RA (Miura et al., 1995). Recently, a new RA-activated pathway, dependent on ATM and cAMP-responsive-element-binding protein (CREB), has been identified that regulates RA-induced neuronal differentiation (Fernandes et al., 2007). Notably, we observed a CREB-binding element (CRE) in the 5′ flanking sequence of the human ZFHX3 promoter and therefore hypothesized that signaling from RA via ATM and CREB activation might play an important role in initiating ZFHX3 gene expression after RA treatment.

We cloned various lengths of the human ZFHX3 promoter (Fig. 1; supplementary material Fig. S1) upstream of the luciferase gene as a reporter (Fig. 1B), transiently transfected these constructs into mouse embryonal carcinoma P19 cells, and measured reporter activity with or without stimulation by increasing amounts of RA. RA activated the neuron-specific 5.6-kb promoter in a dose-dependent manner (Fig. 1B, −5.6), whereas a 5′ 1.6-kb deletion reduced the promoter activity (Fig. 1B, −4.0) to the basal levels observed with 0.42-kb promoters (Fig. 1B, −0.42), which did not respond to RA. A 5′ addition of a 1.6-kb sequence including a CRE consensus sequence to the 0.42-kb basal promoter sequence markedly enhanced the response to RA (Fig. 1B, −5.6Δ). This response was abolished by pre-incubation with the ATM-specific kinase inhibitor KU-55933 (Fig. 1C). To confirm the involvement of CREB, a key survival factor for differentiating neurons, we overexpressed dominant-negative CREB (dn-CREB) by using the pCMV-CREB133 vector, which expressed a mutant variant of the CREB protein in which a serine was mutated to alanine at amino acid 133. This mutation blocks phosphorylation of CREB by protein kinase A, thus preventing cAMP activation of transcription. We observed a significant reduction in transactivation of the RA-responsive promoter construct after RA treatment in cells expressing the dn-CREB (Fig. 1D, −5.6Δ). In addition, mutation of the CREB-binding site in the promoter significantly impaired promoter activation in response to RA with or without expression of dn-CREB (Fig. 1D, −5.6Δmut). These results indicate that ATM and CREB pathway activation is the mechanism employed by RA to transactivate the neuron-specific ZFHX3 promoter.

We expected to observe maximum promoter activation with the 5.6-kb promoter sequence because it contains multiple consensus elements of activators in addition to CRE (see supplementary material Fig. S1). However, the activity of the 5.6-kb promoter was lower than that of a short CRE-containing fragment (5.6Δ), suggesting unknown negative regulatory elements in the promoter sequence.

RA stimulation of P19 cells produced aggregated structures called embryonic bodies on bacterial-grade dishes over 4 days, which were transferred to cell-culture-grade dishes to generate neuron-like cells that adhere to the dishes (Fig. 2B1). We observed suppression of cell adhesion to culture dishes by treatment with KU-55933 (Fig. 2B2). Because there was no significant alteration of activated caspase 3 between cells treated with KU-55933 and control cells (supplementary material Fig. S2), we concluded that the suppression of cell adhesion did not result from an increased number of apoptotic cells but from reduction of adhesive molecules by treatment with KU-55933.

We investigated whether the primary cause of the detachment of cells was related to the function of ATM or ZFHX3 by performing two distinct RNA-silencing experiments, using small interfering RNA (siRNA) targeting (1) Atm RNA and (2) Zfhx3 RNA. Atm siRNA reduced the total amount of ATM protein (Fig. 2A, lane 2) and suppressed Atm mRNA to 30% of the normal levels (Fig. 2E), but still this treatment kept the sufficient levels of activated ATM (pS-ATM; autophosphorylated on Ser1981) (supplementary material Fig. S3B) and there was no effect on adhesive molecules by this treatment (Fig. 2B4,E). By contrast, the suppression of kinase activity with KU-55933 reduced the level of pS-ATM (supplementary material Fig. S3A). The reduction of pS-ATM induced the suppression of ZFHX3, which is associated with decreased levels of adhesive molecules (Fig. 2B2,D). Notably, Atm-siRNA-treated cells stably expressed Zfhx3 and maintained wild-type levels of pS-ATM, and such cells were attached to the culture dish (Fig. 2B4,E). By contrast, suppression of Zfhx3 with KU55933 or direct suppression with Zfhx3 siRNA both induced distinct detachment of cells from culture dishes (Fig. 2B2,B5), which was associated with decreased levels of adhesion molecules (Fig. 2D,F). These results led us to conclude that adhesion suppression by treatment with KU-55933 is essentially the effect of the suppression of Zfhx3. The expression microarray analysis revealed that genes encoding extracellular matrix (ECM) components and various enzyme-linked receptors were significantly suppressed by silencing of Zfhx3 in P19-derived neuron-like cells (supplementary material Table S1). Semiquantitative real-time polymerase chain reaction (RT-PCR) analysis confirmed the significant decrease of expression of genes for adhesion molecules including procollagen type III α1 (Col3a1) and integrin α8 (Itga8), and the platelet-derived growth factor receptor β (Pdgfrb) in ZFHX3-depleted cells (Fig. 2B5,F).

Although Pdgfrb-mutant mice reach adulthood without apparent anatomical defects, their brains are more vulnerable to damage after direct injection of N-methyl-D-aspartate (NMDA) or cryogenic injury (Ishii et al., 2006), indicating that PDGFRB expression is important to protect neurons from glutamatergic excitotoxicity. The association of PDGF receptor and integrin causes receptor clustering, increases PDGF binding and promotes PDGF receptor activation (Zemskov et al., 2009). ZFHX3 might play a key role in inducing both integrin and PDGFRB expression, which would facilitate the associated signal transduction from these membrane receptors.

Oxidative stress is a causal, or at least an ancillary, factor in several adult neurodegenerative disorders, as well as in stroke, trauma and seizures (Coyle and Puttfarcken, 1993). It is considered to be one of the major causes for deficient survival of Purkinje neurons from A-T mutant mice (Chen, P. et al., 2003). Oxidative stress induces PDGFRB-mediated ATM kinase activation in various types of cells (Chen, K. et al., 2003). This newly identified PDGFRB-mediated signal pathway is likely to be an important issue in understanding the mechanism of excitotoxicity. We tested whether PDGFRB contributed to ATM activation in P19-derived neuron-like cells under conditions of oxidative stress induced by hydrogen peroxide. We observed rapid activation of ATM as assessed by its phosphorylation on Ser1981 (pS-ATM) within 15 minutes of exposure of cells to hydrogen peroxide (100 μM) (Fig. 3A) and the activation was increased with increasing dose of hydrogen peroxide (1–500 μM) (Fig. 3B). The activation of ATM in response to oxidative stress was strongly suppressed by AG1433, a specific inhibitor of PDGFRB and of VEGFR-2 (vascular endothelial growth factor receptor 2; also known as FLK-1 and KDR) (Fig. 3C,D). By contrast, the activation of ATM in response to X-ray irradiation was not suppressed by the addition of AG1433 (Fig. 3E,F). Therefore, we concluded that these membrane receptors play an important role in triggering the activation of ATM under oxidative stress in neuronal cells.

Activated ATM is localized to nuclear foci in response to genotoxic stress but displays a diffuse pattern in response to hypotonic stimulation or chloroquine treatment (Bakkenist and Kastan, 2003), or RA stimulation (Fernandes et al., 2007). We observed nuclear foci of activated ATM in response to oxidative-stress-induced DNA DSBs, which colocalized with γ-H2AX, a marker of DNA DSBs (supplementary material Fig. S4). Recently, activation of ATM has been demonstrated in the cytoplasm in response to oxidative stress, which was shown to be independent of its activation in the nucleus by the same agent (Alexander et al., 2010). We failed initially to distinguish the distinct activation of ATM in cytoplasm from that in the nucleus because of the degree of oxidative stress induced (supplementary material Fig. S4). Therefore, we designed more-physiological oxidative stress conditions for neurons to investigate activation of ATM in the cytoplasm independent of that in the nucleus.

Although multiple factors can precipitate oxidative stress in neurons, the neurotransmitter glutamic acid is a major factor that induces excitotoxicity of neurons in the brain. Kainic acid (KA) is a potent glutamate receptor agonist with selectivity towards non-NMDA-type glutamate receptors (Olney et al., 1974). We applied a KA stimulation model to induce metabolic oxidative stress in mouse neurons (Coyle and Puttfarcken, 1993; Zhang et al., 2002). KA stimulation caused two types of seizures in these animals. One case showed repeated bending movement of the head. Another case showed abnormal continuous muscle contraction (supplementary material Movie 1), and induced distinct activation of ATM in the mouse brain at the cerebral cortex, hippocampus and deep cerebellar nuclei (DCN) (supplementary material Fig. S5). We found activated ATM predominantly in the cytoplasm (Fig. 4A1; supplementary material Fig. S6) in the absence of γ-H2AX induction (Fig. 4A2; supplementary material Fig. S6). By contrast, X-ray irradiation induced genotoxic stress and ATM was activated in the nucleus (Fig. 4B1; supplementary material Fig. S6) associated withγ-H2AX foci (Fig. 4B2; supplementary material Fig. S6). We observed distinct vacuolar formation in neurons in DCN by stimulation with KA (supplementary material Fig. S5B3).

We examined in more detail the relationship between ZFHX3 and PDGFRB expression in P19 cells before and after neuronal differentiation. A prominent increase of ZFHX3 and PDGFRB protein expression was observed in the terminally differentiated P19 neuron-like cells at day 7 (Fig. 5A). Screening analysis by in situ hybridization (http://developingmouse.brain-map.org/data/Pdgfrb.html) showed that PDGFRB was highly expressed in DCN. Immunohistochemistry revealed distinct staining of ZFHX3 (Fig. 5C4,C5) and PDGFRB (Fig. 5C6,C7) in large neurons of the DCN of adult control mouse brain, but expression of both proteins was strongly suppressed in ATM-deficient mice (Fig. 5C11–C14). BrdU staining was used in an embryonic rat to distinguish proliferating (Fig. 5B2i) and post-mitotic (Fig. 5B2ii) cells. The correlated expression of ZFHX3 and PDGFRB in the post-mitotic neurons was consistent with the results of the microarray analysis using a model of neuronal differentiation (supplementary material Table S1). Suppression of PDGFRB in DCN of the ATM-deficient mouse might account for the mechanisms of specific neurodegeneration in the cerebellum.

The data described here provide further support for a role for cytoplasmic ATM activation in protection against neuronal cell death. It is well established that a significant amount of ATM protein is present in the cytoplasm of a variety of cell types including neurons (Kuljis et al., 1999; Barlow et al., 1996; Boehrs et al., 2007) (Gorodetsky et al., 2007). More recently Li et al. (Li et al., 2009) showed that cytoplasmic ATM modulates synaptic function. They demonstrated that ATM forms a complex with the synaptic vesicle proteins VAMP2 and synapsin-1 and is responsible for phosphorylation of these proteins. These data suggest that a non-nuclear role for ATM might be important in protecting against neurodegeneration. Alexander et al. (Alexander et al., 2010) have shown that reactive oxygen species rapidly activate ATM in the cytoplasm to repress the kinase mTOR in the mammalian target of rapamycin complex 1 (mTORC1) and induce autophagy. This does not involve signaling from DNA DSBs and the authors make the intriguing suggestion that this activation might involve the oxidation of sulfhydryl groups on ATM to alter its conformation.

In this study, we show that ATM induces ZFHX3 expression during RA-induced neuronal differentiation of P19 cells by activation and binding of CREB to a CRE consensus site located in the ZFHX3 promoter. We also show that ZFHX3 regulates target genes that encode cell adhesion molecules (procollagen type III α1, integrin α8) as well as PDGFRB. PDGFRB is a key regulator not only for connective tissue cells in the induction of adhesion molecules but also for neurons in protection from glutamatergic excitotoxicity (Ishii et al., 2006). Notably, we showed that PDGFRB and/or VEGFR-2 positively supported ATM activation in response to oxidative stress in neurons. We identified significant elevation of Pdgfrb and Vegfr-2 mRNA in P19-derived neuron-like cells (supplementary material Fig. S7). Although neurons are highly susceptible to oxidative stress because of their high rate of oxidative metabolism and low level of antioxidant enzymes (Brooks et al., 2000), membrane receptors might immediately respond to oxidative stress through the activation of ATM in the cytoplasm to protect neurons.

We revealed here the activation of ATM in the cytoplasm in response to excessive excitation of neurons. The excitation of neurons induces metabolic oxidative stress that is a causal, or at least an ancillary, factor of chronic neurodegenerative disorders (Coyle and Puttfarcken, 1993). We showed that PDGFRB and/or VEGFR-2 were required for the activation of ATM in response to oxidative stress. The suppression of PDGFRB in DCN of ATM-deficient mice provides an explanation for the reduction of response to oxidative stress. We observed a high incidence of vacuolar formation specifically in DCN (supplementary material Fig. S5B3), where PDGFRB was highly expressed (Fig. 5C6,C7). Neurons in the DCN are directly innovated by Purkinje cells and most output fibers of the cerebellum originate from these nuclei (supplementary material Fig. S8). Damage of DCN would inevitably induce degeneration of Purkinje cells by loss of direct synaptic connections. We consider the functional unit of Purkinje cells and DCN as important to the etiology of A-T. The distinct activation of ATM in the cytoplasm at the DCN caused by KA stimulation indicated the protective response to oxidative stress induced by the excessive neuronal excitation in these neurons. In particular, DCN responded most strongly to show vacuolar formation. This phenomenon might be related to the fact that the activation of ATM in the cytoplasm induces autophagy via suppression of mTOR (Alexander et al., 2010). We observed intensive accumulation of microtuble-asssociated protein light chain 3 (LC3) in DCN in response to KA treatment (supplementary material Fig. S9). Although the autophagic response was believed to promote acute neuron death because of destruction of organelles through this process (Wang et al., 2008), the study of chronic neurodegenerative disease revealed that autophagy was not the primary cause of neuron death but rather was a protective mechanism for the function of neurons by clearing dysfunctioning organelles. Impaired autophagy is implicated in Parkinson’s disease in the accumulation of dysfunctional mitochondria, leading to neurons loss (Narendra et al., 2008). We should further study the mechanism of the autophagy in response to oxidative stress in relation to the activation of ATM in these neurons.

Expression microarray analysis revealed that ZFHX3 significantly activated most types of procollagen genes except type II, an essential component of the extracellular matrix during wound healing. Because collagen type II is a specific component of cartilage (Cheah et al., 1985), it is reasonable that this type of collagen is not expressed in a model of neuronal differentiation. PDGFRB protein becomes prominent in vessels in the proliferating tissue zone in wounds as early as the first day after surgery (Reuterdahl et al., 1993). Because ZFHX3 regulates procollagen genes as well as PDGFRB, it would be expected to play a crucial role during wound healing. Therefore, malfunction of ZFHX3 might be expected to induce circulatory diseases. Recently, variants in ZFHX3 have been identified that are associated with genetic susceptibility to Kawasaki disease, with increasing risk of aneurismal dilatation in the heart (Burgner et al., 2009). Variants in ZFHX3 are also associated with susceptibility to atrial fibrillation and ischemic stroke (Benjamin et al., 2009; Gudbjartsson et al., 2009). These linkage studies suggest an important role of ZFHX3 in the wound healing process in damaged blood vessels. Notably, PDGFRB is highly expressed in the cerebellum (Lein et al., 2007), the area of the brain that undergoes the specific neurodegeneration that characterizes A-T.

In summary, we propose here three different pathways to trigger the activation of ATM (Fig. 6). The first, signaling from DNA DSBs, induces the activation of ATM at foci in the nucleus. The second, retinoic acid stimulation (Fernandes et al., 2007), activates ATM in a diffuse pattern in the nucleus (Bakkenist and Kastan, 2003). The third, oxidative stress, induces activation of ATM in the cytoplasm (Alexander et al., 2010), which is the pathway in response to neuronal excitation. We found membrane-bound tyrosine kinases are required to trigger the activation of ATM under oxidative stress, which is independent from the response to genotoxic stress. Once activated, ATM can induce autophagy by suppressing mTOR to enhance clearance of damaged organelles. Overall, the observations support a role for cytoplasmic ATM in protecting neurons against oxidative stress and chronic neuronal degeneration.

P19 mouse embryonal carcinoma cells were maintained in α-MEM (Rudnicki and McBurney, 1987). To induce neuronal differentiation, P19 cells were grown as aggregates on bacterial-grade dishes for 4 days with 0.5 μM all-trans retinoic acid (RA) (Sigma, USA) and 10% fetal bovine serum (FBS). Cells were cultured in α-MEM containing 1% FBS without RA on glass slides coated with combinations of 0.01% poly-L-lysine (P8920, Sigma) for 3 hours and 1 μg/ml bovine plasma fibronectin (Invitrogen, USA) overnight at room temperature. Where indicated, the ATM kinase inhibitor KU-55933 [2-morpholin-4-yl-6-thianthren-1-yl-pyran-4-one; 2-morpholino-6-(thianthren-1-yl)-4H-pyran-4-one (Calbiochem, USA)] (10 μM) was added to the culture medium every day, and AG1433 [2-(3,4-dihydroxyphenyl)-6,7-dimethylquinoxaline (Calbiochem, USA)], a specific inhibitor of PDGFRB and VEGFR-2, at the series of concentrations of 0.05, 0.5, 1.0, 5.0 μM was added 30 minutes before adding hydrogen peroxide.

ATM-wild-type and ATM-deficient mice (provided by P. Leder, Harvard Medical School, Boston, MA) were maintained under ambient conditions (23°C, 55% humidity) with controlled light and dark cycles. Genomic DNA was isolated from tail tips and used to amplify a PCR product of the mouse Atm gene by using the following primers: KO-1F, 5′ -TGGTCAGTGTAACAGTCATTGTGC-3′; KO-1R, 5′-AAGGTTGTAGATAGGTCAGCATTG-3′; KO-2R, 5′-AACGAGATCAGCAGCCTCTGTTCC-3′. Primer KO-1F is located 220 bp upstream (intron 34) of exon 34, primer KO-1R is located 123 bp into exon 34, and primer KO- 2R is located 87 bp into the poly(A) region of PGKneo. These primers generate a PCR product of 342 bp from wild-type animals and a targeted PCR product of 406 bp from Atm^−/− mice. Kainic acid [2-carboxy-3-car-boxymethyl-4-isopropenyl-pyrrolidine (Enzo Life Science, USA)] was administered [30 mg/kg, intraperitoneal (i.p.)] to male DDY mice. The whole brain was isolated from 3-month-old mice after perfusion fixation with 4% paraformaldehyde. Pregnant rats (Std Wister ST) were administered 5-bromo-2′-deoxy-uridine (BrdU, 50 mg/kg) for 3 hours to label E14.5 embryos. The embryos were then dissected out, and their brains were fixed in 4% paraformaldehyde and embedded in paraffin.

Cells or mice were X-ray-irradiated at a dose rate of 10 Gy using X-ray machine CAX-210 (Chubu Medical Co. Ltd, Yokkaichi, Japan) operating at 210 kV, 10 mA, for 4 minutes 45 seconds with copper shield.

To generate a series of luciferase reporter plasmids, various lengths of human ZFHX3-promoter regions were cloned into pBluescriptIIKS(+) (Agilent Technologies – Stratagene Products, USA) with SalI and BamHI. The BssHII and BamHI fragments containing ZFHX3 promoter regions were subcloned into the 5′ prime position of the luciferase gene of pGV-B basic vector (Toyo Ink, Japan) between MluI and BglII sites. The CRE site in 1.0 kbp of the ZFHX3 promoter was mutagenized by PCR-based site-directed mutagenesis using a pair of primers, pGV-F (a forward primer with 5′ flanking sequence of luciferase gene on pGV-B): 5′-CAATGTATCTTATGGTACTG-3′, mCRE-R (a reverse direction at CRE site to introduce mutations): 5′-CGGAAATGACCACAGCAAAG-3′, mCRE-F (a forward direction primer at CRE site introduce mutations overlapped with mCRE-F): 5′-CTTTGCTGTGGTCATTTCCG-3′, pGV-R1 (a reverse primer including the first ATG codon of luciferase gene on pGV-B): 5′-CTTTATGTTTTTGGCGTCTTCC-3′, pGV-R2 (a reverse primer at 35-bp downstream from the pGV-R1 in luciferase gene on pGV-B): 5′-CCATCCTCTAGAGGATAGAATG-3′, to mutate the CRE site from 5′-TGACGTCA-3′ to 5′-TGTGGTCA-3′ (supplementary material Fig. S1). The two sets of PCR product were prepared with pGV-F and mCRE-R as a set, and mCRE-F and pVG-R2 as another set. These overlapping primary products were denatured together and reamplified with a new set of primers with pGV-F and pGV-R1. Reamplified products were trimmed with MluI and HindIII, and were subcloned into the same cloning site on the pGV-B basic vector. The CREB dominant-negative expression vector pCMV-CREB133 was from Clontech (USA; Cat. No. 631925).

Sections of 4 μm from 4% paraformaldehyde (PFA)-fixed, paraffin-embedded tissue were used. Tissue sections were deparaffinized and rehydrated and then both those for ZFHX3 (110°C, pressure cooker) and those for ATM (98°C, microwave oven) immunostaining were heated in citrate buffer (0.01 M, pH 6.0). The heat retrieval step was not applied for PDFGRB. Staining was carried out at room temperature. All sections were incubated with methanol containing 0.3% hydrogen peroxide and 1.0% sodium azide to block endogenous peroxidase activity, then incubated with mouse monoclonal anti-ATM Ser1981 phosphorylation site (pS-ATM) monoclonal antibody (at 1:1200 dilution: 7C10D8; Rockland, USA) followed by incubation with horseradish peroxidase (HRP)-conjugated anti-mouse IgG antibody (Envision labeled polymer, Dako Cytomation, Denmark), rabbit polyclonal anti-ZFHX3 antibody (at 1:50 dilution: D1–120; MBL, Nagoya, Japan) followed by HRP-conjugated anti-rabbit IgG antibody (Envision labeled polymer, Dako Cytomation, Denmark) or goat polyclonal anti-mouse PDGFRβ antibody (at 1:100 dilutions: R&D Systems, USA) followed by HRP-conjugated anti-goat IgG antibody [MAX-PO (G), Nichirei, Japan]. Primary antibodies and secondary antibodies were incubated at room for 1 hour. Immunoreactive products were then visualized after adding diaminobenzidine as a chromogen. BrdU labeling and detection was performed using a kit from Roche as per the manufacturer’s recommendations. Tissue sections were counterstained with hematoxylin.

Cells were fixed in 4% PFA in PBS at room temperature for 20 minutes, then washed with 0.25% Triton-X in PBS, and blocked with 10% normal goat serum. Cells were then incubated for 1 hour at room temperature with primary antibodies against rabbit polyclonal anti-ZFHX3 antibody (at 1:100 dilution: D1–120; MBL, Japan), mouse monoclonal anti β-tubulin isotype III (at 1:500 dilution: 3D10; Sigma, USA), monoclonal anti-ATM Ser1981 phosphorylation site (pS-ATM) (at 1:100 dilution: 10H11.E12; Rockland, USA), rabbit polyclonal anti-γH2AX (at 1:100 dilution: BETHYL, USA). After three washes with 0.25% Triton-X in PBS, cells were visualized by secondary antibodies – Alexa-Fluor-488-conjugated goat anti-mouse for mouse monoclonal antibodies and Alexa-Fluor-546-conjugated goat anti-rabbit (at 1:1000 dilution; Molecular Probes, Invitrogen, USA) for rabbit polyclonal antibodies for 1 hour in the dark and, after two washes with PBS, nuclei were stained with 2.0 μg/ml DAPI (at 1:500 dilution of 1 mg/ml solution: 4′,6-diamino-2-phenylindole; Wako, Japan) for 5 minutes.

Cells were washed with washing buffer (10 mM 0.1 M phosphate buffer, 250 mM sucrose, 50 mM NaF), and total cell lysate was prepared with TNE buffer [20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 2 mM EDTA, 1% Nonident-P40, 50 mM NaF]. The proteins were incubated on ice for 15 minutes, and then centrifuged at 15,000 g for 30 minutes at 4°C. The supernatant was obtained and stored at −80°C. The total protein content was measured using the Bradford Assay (Bio-Rad, Hercules, CA). For protein detection, each sample was separated on a 5–20% polyacrylamide gradient gel and the proteins were transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore, Billerica, MA). The membrane was blocked with 3% BSA in Tris-buffered saline containing 0.05% Tween 20 (TBS-T) for 1 hour, washed in TBS-T, and then incubated for 1 hour at room temperature with primary antibodies for mouse monoclonal anti-α-tubulin (at 1:8000 dilution: B-5-1-2; Sigma, USA), rabbit polyclonal anti-ZFHX3 (at 1:2000 dilution: AT-6; MBL, Japan), mouse monoclonal anti-HA-tag (at 1:500 dilution: 5D8; MBL, Japan), mouse monoclonal anti-ATM (at 1:200 dilution: 5C2; Santa Cruz, USA), mouse monoclonal anti-pS-ATM (at 1:1000 dilution: 10H11.E12; Rockland, USA), goat polyclonal anti-PDGFRB (at 1:1000 dilution: R&D Systems, USA) or rabbit polyclonal anti-caspase 3 (at 1:1000 dilution: AAP-113; Assay Designs, USA). After washing in TBS-T, the membrane was incubated with peroxidase-conjugated anti-mouse IgG [at 1:10,000 dilution: anti-mouse IgG (H+L chain)-HRP; MBL, Japan] for mouse monoclonal antibodies, anti-rabbit IgG [at 1:5000 dilution: anti-rabbit IgG (H+L chain)-HRP, MBL, Japan] for rabbit polyclonal antibodies or anti-goat IgG [at 1:5000 dilution: anti-goat IgG (H+L chain)-HRP; MBL, Japan] for a goat polyclonal antibody for 1 hour at room temperature and washed in TBS-T. Immunoreactive signals were visualized by Amersham ECL Plus western blotting detection reagents (GE Healthcare, UK).

Total RNA from cells was isolated using Trizol reagent (Invitrogen, USA), and 1 μg of total RNA was transcribed into cDNA using the Ready-To-Go You-Prime First-Strand Beads kit (GE Healthcare, UK). Real-time PCR was carried out with qPCR MasterMix Plus for SYBR Green (Eurogentec, USA) using the company’s manual procedure: incubation at 95°C for 10 minutes followed by 40 cycles amplification (15 seconds at 95°C, 1 minute at 60°C, 45 seconds at 72°C and 15 seconds at 80°C) for SYBR Green detection. The primers used for real-time measurement of PCR were as follows: Gapdh, 5′-TGTGTCCGTCGTGGATCTGA-3′ and 5′-CCTGCTTCACCACCTTCTTGA-3′; Zfhx3, 5′-TTCTTTTCCTCCTCTCTCCTCATC-3′ and 5′-CGGTCCGTCGGACTTTTG-3′; Col3a1, 5′-GCACAGCAGTCCAACGTAGA-3′ and 5′-TCTCCAAATGGGATCTCTGG-3′); Itga8, 5′-AGTGGGAGGACCTGGAAGTT-3′ and 5′ -AGTGGGAGGACCTGGAAGTT-3′; Pdfgrb, 5′-AACCCCCTTACAGCTGTCCT-3′ and 5′-TAATCCCGTCAGCATCTTCC-3′. The expression of each gene was normalized by the corresponding amount of Gapdh mRNA. The relative amounts of each product were calculated using the comparative CT (2-ΔΔ CT) method described in User Bulletin #2 of the ABI Prism 7500 fast Sequence Detection System (Applied Biosystems, USA).

RNA interference (RNAi) was performed using Stealth/siRNA (Invitrogen, USA) duplex oligoribonucleotides against Zfhx3 (ATBF1) (5′-UACACUGGUCAGACCACUGUCCUUG-3′ and 5′-CAAGGACAGUGGUCUGACCAGUGUA-3′), and three sets of duplex oligoribonucleotides against Atm (5′-UGAACUUCCCGAUAAUCCACAAGG-3′ and 5′ -CCCUUGUGGAUUAUCAGGAAGUUCA-3′; 5′ -UAAACAGAGAGAUACUUUCUCCUGC-3′ and 5′ -GCAGGAGAAAGUAUCUCUCUGUUUA-3′;5′-UUAGAAGGCCCACUUCCUCUUUGGC-3′ and 5′-GCCAAAGAGGAAGUGGGCCUUCUAA-3′). The transfection was performed with Lipofectamine RNAiMAX four times to P19 cells during the neuronal differentiation process on days 0, 1, 3 and 5. Stealth RNAi negative control duplex (medium GC duplex) was also transfected four times, following the protocol of neuronal differentiation of P19 cells (Rudnicki and McBurney, 1987).

Statistical analysis of results from luciferase assays and western blotting was performed using the Mann-Whitney U-test with Bonferroni Correction. All values were depicted as mean ± s.e.m. from at least three independent experiments and considered significant if *P<0.01. All the results from RT-PCR were normalized by the expression levels of Gapdh and compared as the increased ratio from control levels. Statistical significance was assessed with Student’s t-test: **P<0.001.

We thank Yuji Fujinawa, Osamu Yamamoto and John Luff for excellent technical assistance, and Masaaki Inoue and Koichi Tsuneyama for scientific discussions. This study was supported by: a Grant-in-Aid for Special Research from Nagoya City University (Y.M.); a Grant-in-Aid for Exploratory Research (Y.M.); a Grant-in-Aid for Scientific Research [C] (M.K.) from Ministry of Education, Culture, Sports, Science and Technology of Japan; a Grant-in-Aid from The New Energy and Industrial Technology Development Organization of Japan (Y.M.); and a Grant-in-Aid from Japan Science and Technology Agency (Y.M.).