Article - Cell Press

Molecular Cell

Article DNA-Independent PARP-1 Activation by Phosphorylated ERK2 Increases Elk1 Activity: A Link to Histone Acetylation Malka Cohen-Armon,1,* Leonid Visochek,1 Dana Rozensal,1 Adi Kalal,1 Ilona Geistrikh,1,3 Rodika Klein,1 Sarit Bendetz-Nezer,2 Zhong Yao,2 and Rony Seger2 1

Neufeld Cardiac Research Institute, Sackler School of Medicine, Tel-Aviv University, Tel-Aviv 69978 Department of Biological Regulation, The Weizmann Institute of Science, Rehovot 76100 3 Life Sciences, Bar-Ilan University, Ramat Gan 52900, Israel *Correspondence: [email protected] DOI 10.1016/j.molcel.2006.12.012 2

SUMMARY

PolyADP-ribose polymerases (PARPs) catalyze a posttranslational modification of nuclear proteins by polyADP-ribosylation. The catalytic activity of the abundant nuclear protein PARP-1 is stimulated by DNA strand breaks, and PARP-1 activation is required for initiation of DNA repair. Here we show that PARP-1 also acts within extracellular signal-regulated kinase (ERK) signaling cascade that mediates growth and differentiation. The findings reveal an alternative mode of PARP-1 activation, which does not involve binding to DNA or DNA damage. In a cell-free system, recombinant PARP-1 was intensively activated and thereby polyADP-ribosylated by a direct interaction with phosphorylated ERK2, and the activated PARP-1 dramatically increased ERK2-catalyzed phosphorylation of the transcription factor Elk1. In cortical neurons treated with nerve growth factors and in stimulated cardiomyocytes, PARP-1 activation enhanced ERK-induced Elk1-phosphorylation, core histone acetylation, and transcription of the Elk1-target gene c-fos. These findings constitute evidence for PARP-1 activity within the ERK signal-transduction pathway. INTRODUCTION PolyADP-ribosylation is a rapid and transient posttranslational modification of nuclear proteins affecting the interaction between proteins in the chromatin and the binding of proteins to DNA (Schreiber et al., 2006). The reaction is catalyzed by polyADP-ribose polymerases (PARPs). It is initiated by the transfer of an ADP-ribosyl moiety from NAD to glutamate or aspartate residues of target proteins, and it proceeds by the successive transfer and polymerization of ADP riboses in the target protein, rapidly creating

long branched polymers that are subsequently degraded by polyADP-ribose glycohydrolase (PARG) (Schreiber et al., 2006; Lautier et al., 1993). PARP-1, the most abundant PARP, is activated and thereby auto-polyADP-ribosylated by binding to DNA strand breaks (Schreiber et al., 2006). PolyADP-ribosylation of PARP-1 substrates in the chromatin, mainly of its prominent substrate linker histone H1, causes loosening of the highly condensed chromatin structure, rendering the DNA accessible to transcription and repair enzymes (Kraus and Lis, 2003; Tulin and Spradling, 2003; Rouleau et al., 2004). PolyADP-ribosylated PARP-1 also participates in the recruitment of proteins that regulate transcription or DNA repair (Schreiber et al., 2006). We recently showed that PARP-1 in brain cortical neurons is rapidly activated in the absence of DNA damage, downstream of extranuclear signal transduction patterns mediated by phospholipase C (PLC) (Homburg et al., 2000; Visochek et al., 2005). Extracellular signal-regulated kinases (ERKs) are key transmitters of signals mediated by PLC activation (Huang and Reichardt, 2001). These kinases operate within intracellular signal transduction pathways of mitogen-activated protein kinases (MAPKs) and regulate a variety of cell functions, including proliferation, growth, and differentiation (Seger and Krebs, 1995; Huang and Reichardt, 2001; Yoon and Seger, 2006). ERK1 and ERK2, once phosphorylated by the MAPK kinase MEK1/2, migrate into the nucleus where they phosphorylate and activate numerous substrates, including transcription factors (Seger and Krebs, 1995; Lenormand et al., 1998; Pouyssegur and Lenormand, 2003; Yoon and Seger, 2006). Because similar signal transduction patterns mediate the activation of PARP-1 and of ERK1 and ERK2, we investigated the possibility of interplay between the activated proteins. Our findings disclosed an alternative mode of PARP-1 activation by phosphorylated ERK2 in the absence of DNA damage, which operates within the ERK signaling cascade mediating growth and differentiation. Activated PARP-1 increased ERK2-catalyzed Elk1 phosphorylation, histone acetylation, and the expression of the Elk1-targeted gene c-fos.

Molecular Cell 25, 297–308, January 26, 2007 ª2007 Elsevier Inc. 297

Molecular Cell PARP-1 Acts Within ERK Signaling Pathway

RESULTS Evidence for Involvement of the ERK Cascade in Activation of PARP-1 The ERK-signaling cascade is activated downstream of PLC in neurons treated with nerve growth factors (Huang and Reichardt, 2001) and in contracting cardiomyocytes treated with angiotensin II (Ang II; Yamazaki et al., 1999). PARP-1 was also activated in these treated cells (Figure 1A and see Figures S1 and S5 in the Supplemental Data available with this article online). Interestingly, MEK inhibitors (U0126 and PD98059) that prevent ERK phosphorylation (Shaul and Seger, 2004) (Figure 1B) suppressed the activation of PARP-1 (Figure 1A and Figures S1 and S5). PARP-1 activation was measured by the shift in its isoelectric point toward a lower pH, due to polyADP-ribosylation (Figure 1A; Experimental Procedures). PolyADP-ribosylated proteins were immunodetected in the treated cells (Figures S1 and S5). In view of the effect of MEK inhibitors on PARP-1 activation, it was of interest to investigate a possible effect of phosphorylated ERKs (pERKs). PARP-1 Activation by Recombinant pERK2 in Isolated Nuclei and Nuclear Extracts In these experiments, we used a semi-in vitro system in which glutathione S-transferase (GST)-bound recombinant pERK2 (r-pERK2) was inserted into digitonin-permeabilized brain cortical neurons in primary culture and into isolated nuclei of cortical slices (Figure 2; Experimental Procedures). An increase in pERK2 in the treated cells was accompanied by a dramatic increase in polyADPribosylated nuclear proteins (Figure 2A) and PARP-1 polyADP-ribosylation (Figure 2B). Phosphatase activity interfered with PARP-1 activation (Figure 2B). Also, PARP-1 was not activated when r-pERK2 was replaced by an inactive, nonphosphorylated ERK2 construct D25-ERK2 (Eblen et al., 2001; Figure 2B). These results suggested that r-pERK2 induced PARP-1 activation in the nuclei. Lack of DNA damage in the permeabilized neurons excluded a possible PARP-1 activation by DNA breaks rather than by r-pERK2 insertion (Figure S2). Possible activation of endogenous PARP-1 by r-pERK2 was also examined in nuclear protein extracts. PARP-1 was highly [32P]polyADP-ribosylated following application of r-pERK2 and [32P]NAD (1000 Ci/mmol; 1 mCi/sample; 20 nM) to nuclear protein extracts prepared from cortical slices (Figure 3A). GST alone did not induce PARP-1 activation (Figure 3A). Furthermore, neither the GST-bound inactive nonphosphorylated ERK2 construct (D25-ERK2) nor the GST-bound nonphosphorylated ERK2 construct (ERK2) induced activation of PARP-1 (Figure 3A). Thus, PARP-1 in the nuclear protein extracts was evidently activated by the applied r-pERK2. The possibility of pERK2-induced PARG inhibition was excluded (Figure S3, Supplemental Data), indicating that the pERK2-induced polyADP-ribosylation of PARP-1 did not result from inhibition of PARG activity.

Figure 1. MEK Inhibitors Interfere with PARP-1 Activation (A) PARP-1 activation was measured by the polyADP-ribosylationinduced shift in its isoelectric point toward lower pH values. Rat brain cortical slices were incubated with NGF (100 ng/ml, 10 min, 95%O2/ 5%CO2, 25 C), before or after treatment with the PARP-1 inhibitor 3-AB (0.5 mM), MEK inhibitors PD98059 (50 mM), or U0126 (10 mM). Nuclear proteins were extracted, separated by two-dimensional gel electrophoresis, transferred to nitrocellulose membranes (western blots), and immunolabeled for PARP-1 (n = 3). (B) Phosphorylated ERK1 and ERK2 in nuclear extracts of brain cortical neurons treated with NGF or BDNF (100 ng/ml, 10 min, 37 C). Treatment with U0126 (10 mM) suppressed their phosphorylation, as indicated by immunolabeling (n = 2).

Activation of Purified PARP-1 and Recombinant PARP-1 by Recombinant pERK2 We examined the effect of r-pERK2 on both purified PARP-1 and recombinant PARP-1 (r-PARP-1). Incubation of PARP-1 with r-pERK2 in the presence of [32P]NAD resulted in enhanced [32P]polyADP-ribosylation of PARP-1 (Figures 3B–3D). Interestingly, whereas [32P]polyADPribosylation of r-PARP-1 by incubation with r-pERK2 was substantial, it was hardly achieved by incubation with degraded DNA (Figure 3D). Furthermore, the effect of r-pERK2 was not limited to automodification of PARP-1; the prominent PARP-1 substrate, linker histone H1 (Buki et al., 1995), was also [32P]polyADP-ribosylated following incubation with r-pERK2, PARP-1, and [32P]NAD (Figure 3B). However, the pERK2-induced PARP-1 polyADPribosylation was attenuated in the presence of H1 (Figure 3B), possibly because r-pERK2 and histone H1 compete for mutual binding domains in PARP-1.

298 Molecular Cell 25, 297–308, January 26, 2007 ª2007 Elsevier Inc.


Figure 2. Activation of Endogenous PARP-1 by Insertion of Recombinant pERK2 into Isolated Nuclei (A) r-pERK2 was inserted into digitonin-permeabilized (40 mg/ml digitonin, 5 min on ice) cultured cortical neurons treated with phosphatase inhibitors. PolyADP-ribosylated proteins were immunolabeled before and after r-pERK2 insertion. Confocal images show enhanced immunolabeling of pERKs (red) and polyADP-ribosylated proteins (PAR, green) in the nuclei after r-pERK2 insertion. Merged labeling (yellow) may indicate colocalization of polyADP-ribosylated proteins and pERKs (n = 5). (B) PolyADP-ribosylation of PARP-1 was tested before and after insertion of r-pERK2 into nuclei isolated from cortical slices treated with phosphatase inhibitors (PI; 300 nM okadaic acid and 1 mM sodium orthovanadate) and protease inhibitors. PolyADP-ribosylation was assayed by the shift in the isoelectric point of PARP-1 toward a lower pH. The isoelectric point of PARP-1 was shifted neither after r-pERK2 insertion into nuclei prepared from slices treated with 3-AB (0.5 mM) nor after insertion of a nonphosphorylated, inactive ERK2 construct (D25-ERK2) instead of r-pERK2 (n = 5).

Given that only some 60%–70% of the r-pERK2 in the samples was phosphorylated (see Figure 4A), the enhanced polyADP-ribosylation of PARP-1 was evidently obtained by incubation with r-pERK2 at a ratio of about one molecule of PARP-1 to two molecules of r-pERK2 (Figure 3C). This ratio is not compatible with the stoichiometry of an enzymatic reaction, suggesting that the kinase activity of pERK2 was not involved in PARP-1 activation. Effect of DNA on the Activation of r-PARP-1 by r-pERK2 Because PARP-1 is a DNA-binding protein, known to be activated by binding to nicked DNA (Schreiber et al., 2006), we examined the effect of exogenous DNA on pERK2-induced PARP-1 activation in a cell-free system. Nicked DNA (DNA pretreated with DNase-I), sheared DNA (ssDNA) (salmon sperm DNA), and intact genomic DNA (gDNA) (Experimental Procedures) were tested for

their effects on pERK2-induced polyADP-ribosylation of PARP-1. Incubation of r-PARP-1 with r-pERK2 and [32P]NAD resulted in its substantial [32P]polyADP-ribosylation (Figure 3D). Intact genomic DNA or sheared DNA did not interfere with the pERK2-induced PARP-1 polyADP-ribosylation (Figure 3D). However, it was completely blocked by the application of nicked DNA (Figure 3D), suggesting that the binding of r-PARP-1 to breaks in DNA (Menissierde Murcia et al., 1989) precludes its interaction with r-pERK2. Despite the inhibitory effect of nicked DNA on pERK2induced PARP-1 activation, r-PARP-1 was hardly polyADP-ribosylated by the nicked DNA (Figure 3D). These findings exclude the possibility of [32P]polyADP-ribosylation of PARP-1 in our experiments by residual quantities of nicked DNA, which theoretically could reside in the samples of recombinant ERK proteins.



Figure 3. Activation of PARP-1 by Recombinant pERK2 in Nuclear Protein Extracts and in Cell-Free Systems (A) r-pERK2 enhances [32P]polyADP-ribosylation of endogenous PARP-1 in nuclear protein extracts of cortical neurons treated with phosphatase inhibitors (PI). (Upper panel) Autoradiograms of [32P]polyADP-ribosylated proteins in nuclear protein extracts following incubation (20 min, 30 C) with [32P]NAD (lane 1, control) and with GST (200 ng/sample; lane 2) or 100 ng/sample r-ERK2 (ERK2; lane 7), r-pERK2 (pERK2; lanes 3, 4, and 6), or inactive, nonphosphorylated ERK2 construct (D25-ERK2; lane 5). For comparison, r-pERK2 was applied in the absence of phosphatase inhibitors (lane 4) or in the presence of PARP inhibitor (3-AB; 0.5 mM; lane 6). (Lower panels) Immunolabeled PARP-1 and pERK2 in the samples (n = 5). (B) Purified PARP-1 was [32P]polyADP-ribosylated by incubation with r-pERK2 and [32P]NAD. (Upper panel) Autoradiograms of [32P]polyADP-ribosylated PARP-1 (100 ng/ sample) following incubation (20 min, 30 C) with r-pERK2 (100 ng/sample) and [32P]NAD (1 mCi/sample; 20 nM). Recombinant histone H1 (200 ng/sample; Upstate) and DNasetreated nicked DNA (nDNA; 1 mg/sample) were added to the incubation mixture, as indicated. Lower panels display immunolabeled PARP-1, pERK2, and H1 in the samples (n = 3). (C) Dose-dependent [32P]polyADP-ribosylation of purified PARP-1 incubated with increasing concentrations of r-pERK2. (Upper panel) Autoradiograms show [32P]polyADP-ribosylated PARP-1 following incubation (30 C) with [32P]NAD (1 mCi/sample; 20 nM) and r-pERK2. PARP-1 (50 ng/sample) (lanes 1 and 8) was incubated for 20 min (lanes 2–7) or for 5 min (lanes 9–13) with r-pERK2 (0.1, 1, 5, 10, and 50 ng/sample, respectively, in each case). The effect of r-pERK2 at 50 ng/sample was completely suppressed by 3-AB (0.5 mM) (lane 7). (Lower panels) Immunolabeled PARP-1 and pERK2 in the samples (n = 2). (D) Effect of intact, sheared, and nicked DNA on pERK2-induced [32P]polyADP-ribosylation of recombinant PARP-1 (r-PARP-1). (Upper panel) Autoradiograms of [32P]polyADP-ribosylated r-PARP-1 (100 ng/sample) following incubation (20 min, 30 C) with [32P]NAD (1 mCi/ sample; 20 nM) and r-pERK2 (100 ng/sample) in the absence or presence of DNA (1 mg/sample of genomic DNA [gDNA], sheared DNA [ssDNA], or DNase-treated nicked DNA [nDNA], as indicated). (Lower panels) Immunolabeled PARP-1 and pERK2 (labeled for pERK2 and ERK2) in the samples (n = 3 for each treatment).



Characteristics of pERK2-Induced PARP-1 Activation To further examine PARP-1 activation by r-pERK2, we tested the following ERK2 constructs for their ability to activate r-PARP-1: wild-type ERK2 (ERK2); K54A(KA)-ERK2 (KA-ERK2), which lacks kinase activity but is readily phosphorylated by MEKs (Seger et al., 1992); and 316A-ERK2 (CRS-ERK2), in which the acidic amino acids of the docking domain (CRS/CD) (Yoon and Seger, 2006) are replaced by alanines, reducing its ability to interact with proteins (Wolf et al., 2001; Yoon and Seger, 2006). Phosphorylated ERK2 and its mutants (prepared by coexpression with active MEK in bacteria; Jaaro et al., 1997) and their nonphosphorylated counterparts were incubated with r-PARP-1 and [32P]NAD. PARP-1 activation was assayed by measuring its [32P]polyADP-ribosylation (Figure 4A). r-pERK2 and its mutants substantially activated r-PARP-1, whereas nonphosphorylated r-ERK2 and its mutants did not (Figure 4A). Also, r-PARP-1 was substantially activated in the presence of the inactive but phosphorylated r-KA-pERK2, but not by incubation with the nonphosphorylated mutant r-KA-ERK2 (Figure 4A). r-PARP-1 was also activated by the phosphorylated CRS-pERK2 construct (Figure 4A), but the polyADP-ribosylation of r-PARP-1 obtained in this case was lower than that obtained with the other pERKs, suggesting a possible involvement of the docking CRS/CD region of ERK2 in the interaction between pERK2 and PARP-1. The nonphosphorylated CRS-ERK2 did not activate r-PARP-1 (Figure 4A). To summarize, the nonphosphorylated constructs did not induce PARP-1 activation, yet ATP was not required for activation of r-PARP-1 by the phosphorylated ERK2 constructs, and r-PARP-1 was activated by the inactive but phosphorylated KA-pERK2 construct. These findings clearly indicate that the phosphorylation of r-pERK2, rather than its kinase activity, was required for activation of r-PARP-1. In addition, r-PARP-1 was scarcely phosphorylated by r-pERK2 (Figure 4B). An interesting observation was the [32P]ADP labeling of r-pERK2 constructs in the protein mixtures containing [32P]NAD and r-PARP-1. This was probably attributable to PARP-1 activity and indicates an interaction between PARP-1 and ERK2 in the protein mixtures (Figure 4A). The enzymatic activity of the pERK2-activated r-PARP-1 was compared with that of r-PARP-1 activated by nicked DNA (Figures 4C and 4D and Supplemental Data). At the low range of [32P]NAD concentrations, r-PARP-1 was highly [32P]polyADP-ribosylated by incubation with r-pERK2, whereas r-PARP-1 incubated with nicked DNA was hardly [32P]polyADP-ribosylated at all (Figures 4C). This may imply a higher affinity of pERK2-activated PARP-1 for NAD than the affinity of PARP-1 activated by nicked DNA. This notion was supported by the calculated affinity constant (KM) of 1.1 ± 0.2 mM for pERK2-activated r-PARP-1, about 50 times lower than the reported KM of r-PARP-1 activated by nicked DNA (59 mM; Mendoza-

Alvarez and Alvarez-Gonzalez, 1993). The KM for pERK2activated r-PARP-1 was calculated using the MichaelisMenten formulation for a competition between two substrates for a common binding site (i.e., binding of NAD and [32P]NAD to pERK2-activated PARP-1; Figure 4D and Supplemental Data). Direct Interaction between PARP-1 and pERK2 Possible interaction between PARP-1 and pERK2 was examined by coimmunoprecipitation in mixtures containing PARP-1, pERK2, and ERK2 constructs. Antibodies directed against ERK2 caused coimmunoprecipitation of r-PARP-1 with r-ERK2 or with r-pERK2 (Figure 5A). No crossreactivity was detectable between PARP-1 and the anti-ERK2 antibody (Figure 5A, lane 3). PARP-1 also failed to coimmunoprecipitate with GST (Figure 5A, lane 1), excluding a possible role for GST in the interaction of r-pERK2 or r-ERK2 constructs with r-PARP-1 (see also Figure 3A). Endogenous PARP-1 also coimmunoprecipitated with endogenous ERK2, either phosphorylated or nonphosphorylated in nuclear protein extracts of brain cortical neurons treated with phosphatase inhibitors (Figure 5B). Furthermore, when endogenous PARP-1 and ERK2 were both activated by treatment of brain cortical slices with nerve growth factors (Figure 1 and Figure S1), coimmunoprecipitation of PARP-1 with ERK2 was more efficient (Figure 5B). These findings suggest a possible direct interaction between polyADP-ribosylated PARP-1 and phosphorylated ERK2 in the nuclear extracts. This possibility was further tested by using the dot blot technique (Experimental Procedures). In this technique, binding of pERK2 to polyADP-ribosylated endogenous PARP-1 was detected by the binding of pERK2 to antibody directed against PARP-1, which was applied in dots on nitrocellulose membranes (Supplemental Data). Binding of endogenous ERK2 and its substrate transcription factor Elk1 to the applied PARP-1 antibody suggested that ERK2 and Elk1 interacted with PARP-1 in nuclear protein extracts prepared from brain cortical slices treated with NGF (Figure 5C), when PARP-1 was polyADP-ribosylated (Figure 1A and Figure S1) and ERK2 was phosphorylated (Figure 1B). We therefore examined the possibility of exclusive interaction between the recombinant proteins polyADP-ribosylated PARP-1, pERK2, and Elk1. r-PARP-1 was polyADP-ribosylated by incubation with sheared DNA and b-NAD. Equal amounts of PARP-1 and polyADP-ribosylated PARP-1 were applied in dots on nitrocellulose membranes, and the binding of r-pERK2, r-ERK2, and r-Elk1 (that lacks the DNA-binding domain of Elk1; Supplemental Data) to r-PARP-1, either unmodified or polyADP-ribosylated, was examined. We found that r-ERK2, r-pERK2, and r-Elk1 bound preferentially to polyADP-ribosylated r-PARP-1 (Figure 5D). In the absence of r-ERK2 or r-pERK2, r-Elk1 did not bind to r-PARP-1, whether polyADP-ribosylated or not (Figure 5D), implying a possible binding of Elk1 to PARP-1 via ERK2. Nonspecific binding of ERK2 and Elk1 to PARP-1 was excluded



Figure 4. Characteristics of pERK2-Induced PARP-1 Activation (A) r-PARP-1 was polyADP-ribosylated by incubation with pERK2 constructs. (Upper panel) Autoradiograms of [32P]polyADP-ribosylated r-PARP-1 (100 ng/sample) after incubation (20 min, 30 C) with [32P]NAD (1 mCi/sample, 20 nM) and recombinants of phosphorylated ERK2 or ERK2, as indicated. Lane 1, PARP-1 incubated with [32P]NAD only. Lanes 2, 4, and 6, r-PARP-1 incubated with [32P]NAD and pERK2 (100 ng/sample; lane 2), CRS-pERK2 (150 ng/sample; lane 4), and KA-pERK2 (150 ng/sample; lane 6). Lanes 3, 5, and 7, r-PARP-1 incubated with [32P]NAD and the respective recombinants of nonphosphorylated ERK2 (100 ng/sample ERK2, 200 ng/sample CRS-ERK2 and 200 ng/sample KA-ERK2). (Lower panels) Immunolabeled PARP-1, pERK2, and ERK2 in each sample (n = 3). (B) r-PARP-1 was hardly phosphorylated by r-pERK2. Autoradiograms present pERK2-catalyzed [32P]phosphorylation of r-PARP-1 (1 mg/sample) and r-Elk1 (600 ng/sample) following incubation (20 min, 30 C) with r-pERK2 (100 ng/sample), 0.1 mM ATP, [32P]ATP (3000 Ci/mmol, 10 mCi/sample), and phosphatase inhibitors. [32P]phosphorylated proteins were autoradiographed (upper panel) and immunolabeled (lower panels), as indicated (n = 4). (C) [32P]polyADP-ribosylation of r-PARP-1 activated by r-pERK2 compared with [32P]polyADP-ribosylation of r-PARP-1 activated by nicked DNA. (Left) r-PARP-1 (100 ng/sample) was incubated (10 min, 30 C) with [32P]NAD (10–100 nM; 1000 Ci/mmol) in the absence (control, ) or in the presence of r-pERK2 (pERK2; 100 ng/sample, ,) or nicked DNA (nDNA; 800 ng/sample, A). In both treatments, r-PARP-1 was dose-dependently [32P]polyADP-ribosylated. Autoradiograms of [32P]polyADP-ribosylated PARP-1 (upper panel) were scanned and quantitated (lower panel). r-PARP-1 and r-pERK were immunolabeled in the samples (n = 4). (Right) r-PARP-1 (100 ng/sample) was dose-dependently [32P]polyADP-ribosylated (10 min, 30 C) at a wider range of NAD concentrations (a mixture of [32P]NAD and b-NAD at the indicated concentrations) in the presence of r-pERK2 (pERK2; 100 ng/sample) or nicked DNA (nDNA; 800 ng/sample). Autoradiograms show [32P]polyADP-ribosylated PARP-1. r-PARP-1 and r-pERK were immunolabeled in the samples (n = 3; Supplemental Data). (D) Dose-dependent polyADP-ribosylation of r-PARP-1 (100 ng/sample), activated by incubation with r-pERK2 (100 ng/sample; 10 min, 30 C) at a constant [32P]NAD concentration (20 nM) and increasing concentrations of b-NAD (1 nM to 200 mM). Autoradiograms of [32P]polyADP-ribosylated r-PARP-1 were quantitated after scanning. The concentration of b-NAD causing 50% decay in the [32P]polyADP-ribosylation of r-PARP-1 is indicated. PARP-1 and pERK were immunolabeled in the samples (n = 3; Supplemental Data).



Figure 5. Evidence for a Direct Interaction between PARP-1 and ERK2 (A) Coimmunoprecipitation of r-PARP-1 (500 ng/sample) with r-pERK2 or r-ERK2 (1.5 mg/sample) by antibody directed against the C-terminal of ERK2. Lane 1, control, coimmunoprecipitation of r-PARP-1 with GST (2 mg/sample). Lane 2, control, nonspecific binding of r-PARP-1 to protein A Sepharose beads. Lane 3, control, nonspecific immunoprecipitation of r-PARP-1 with the antibody directed against ERK2. Lanes 4 and 5, coimmunoprecipitation of r-PARP-1 with r-ERK2 and r-pERK2, respectively (n = 3). (B) Coimmunoprecipitation of PARP-1 with ERK2 in nuclear protein extracts. (Left panel) Protein extracts were prepared from brain cortical slices treated with phosphatase inhibitors (okadaic acid 300 nM and orthovanadate 1 mM; 30 min 95%O2/5%CO2, 25 C). Lanes 1 and 2, PARP-1, ERK1/2, and pERK1/2 immunolabeled in the input sample. Lane 3, control, nonspecific binding of PARP-1 to protein A Sepharose beads. Lane 4, coimmunoprecipitation of endogenous PARP-1 with endogenous ERK2 and pERK2 by antibody directed against the C terminus of ERK2. (Right panel) Coimmunoprecipitation of PARP-1 with ERK2 in nuclear protein extracts prepared from brain cortical slices treated with NGF or BDNF (100 ng/ml, 10 min, 95%O2/5%CO2, 25 C) applied in the absence or presence of MEK inhibitor PD-98059 (50 mM), as indicated (n = 4). (C) Dot blots, used to test the binding of endogenous ERK2 and Elk1 with endogenous PARP-1 by protein binding to anti-PARP-1 antibody, applied in dots on nitrocellulose membranes. Membranes were incubated (1 hr, 25 C) with nuclear protein extracts prepared from brain cortical slices, untreated or treated with BDNF (100 ng/ml, 10 min, 95%O2/5%CO2, 25 C) and PARP-1 inhibitor (PJ-34; 20 mM) or MEK inhibitor (U0126; 10 mM), as indicated. Proteins bound to the applied anti-PARP-1 antibody were immunolabeled for ERK2 and Elk1, as indicated (n = 5; Supplemental Data). (D) Binding of pERK2, ERK2, and Elk1 constructs to r-PARP-1, either unmodified or polyADP-ribosylated. r-PARP-1 (200 ng/sample) was polyADPribosylated by incubation (30 C, 20 min) with b-NAD (0.2 mM) and sheared DNA (DNA; 1 mg/sample). PolyADP-ribosylation of r-PARP-1 was verified by immunolabeling with antibody directed against ADP-ribose polymers (PAR). Equal amounts of r-PARP-1 and polyADP-ribosylated r-PARP-1 were applied in dots on nitrocellulose membranes. Membranes were exposed (1 hr, 25 C) to r- ERK2, r-pERK2, r-Elk1 (20 mg/ml) and GST (30 mg/ml; control). Binding of the recombinant proteins (left panel, ERK2 and pERK2; right panel, Elk1) to r-PARP-1 and polyADP-ribosylated r-PARP-1 was detected by immunolabeling, as indicated. Neither r-pERK2 nor r-Elk1 bound to the DNA in the samples (DNA was detected with 1 mg/ml ethidium bromide under UV illumination) (n = 3 for each treatment; Supplemental Data).

(Figures 5C and 5D and Supplemental Data). In addition, results of experiments performed in collaboration with Professor Alexander Bu¨rkle (Konstanz University) failed to provide any evidence for binding of r-pERK2 or r-ERK2 to the ADP-ribose polymers of polyADP-ribosylated PARP-1 (data not shown). Taken together, these results pointed to a direct interaction between pERK2 and PARP-1 and preferentially with polyADP-ribosylated PARP-1. PolyADP-Ribosylated PARP-1 Increases pERK-Catalyzed Elk1 Phosphorylation By using a cell-free system, we examined whether the direct interaction between polyADP-ribosylated r-PARP-1, r-pERK2, and r-Elk1 (Figure 5D) affects their activity. The r-pERK2-catalyzed Elk1 phosphorylation was dramatically enhanced in the presence of polyADP-ribosylated

r-PARP-1 (Figure 6A). In these experiments, PARP-1 was polyADP-ribosylated by r-pERK2 in the presence of b-NAD. When r-Elk1 and ATP were added to the mixture of polyADP-ribosylated PARP-1 and r-pERK2, r-Elk1 was much more abundantly phosphorylated than r-Elk1 phosphorylated by r-pERK2 alone (Figure 6A, compare lanes 1 and 2). Also, pERK2-catalyzed r-Elk1 phosphorylation was much more efficient in the presence of polyADP-ribosylated PARP-1 than in the presence of nonactivated PARP-1 (Figure 6A, compare lanes 2 and 4). Thus, interaction between r-PARP-1 and r-pERK2, yielding polyADP-ribosylation of r-PARP-1, enhanced pERK2-catalyzed phosphorylation of r-Elk1 (Figures 6A, 4A, and 4C). We also examined the effect of PARP-1 activation on Elk1 phosphorylation in cortical neurons treated with nerve growth factors and in cardiomyocytes treated with Ang II. ERKs and Elk1 were both phosphorylated in the



Figure 6. PolyADP-Ribosylated PARP-1 Increases pERK-Catalyzed Elk1 Phosphorylation and Expression of Elk1-Target Gene c-fos (A) Effect of polyADP-ribosylated r-PARP-1 on phosphorylation of r-Elk1 by r-pERK2. r-Elk1 (600 ng/sample) was added to a mixture of r-pERK2 (100 ng/sample) and r-PARP-1 or polyADP-ribosylated r-PARP-1 (100 ng/sample). r-PARP-1 was polyADP-ribosylated by incubation with r-pERK2 (10 min, 30 C) and b-NAD (0.2 mM). Elk1 was phosphorylated by incubation with r-pERK2 and ATP (2 mM) (20 min, 30 C) in the absence (lanes 1 and 3) or in the presence of PARP-1, either polyADP-ribosylated (lane 2) or not (lane 4); r-Elk1 and ATP were added 15 min after (lane 2) or before (lane 4) the addition of b-NAD. PolyADP-ribosylated PARP-1 is immunolabeled by anti-PAR and PARP-1 antibodies. pERK2 and pElk1 constructs are immunolabeled, as indicated (n = 6). (B) Effect of PARP-1 activation on Elk1 phosphorylation and c-fos expression in cortical neurons treated with NGF (100 ng/ml, 7 min, 37 C), before or after PARP-1 inhibition by PJ-34 (20 mM, 30 min, 37 C). Phosphorylation of ERK1, ERK2, and Elk1 and the levels of PARP-1 and c-fos protein in the nuclear protein extracts, prepared 30, 60, and 120 min after treatment with NGF, were detected by immunolabeling and quantitated after scanning. Changes in levels of phosphorylated Elk1 (pElk1) and c-fos are displayed in the bar diagrams (n = 3). (C) (Top) PARP-1 activation increases the expression of c-fos gene in rat brain cortical neurons treated with NGF (100 ng/ml, 7 min, 37 C). Expression levels of the c-fos gene were measured 15, 30, and 60 min after treatment by q-PCR relative to the expression levels of two reference genes, b-actin and GAPDH (Experimental Procedures). NGF-induced expression of c-fos gene (A) was suppressed in cells treated for 30 min with either PARP-1 or MEK inhibitors (PJ-34, 10 mM [gray square] or U0126, 10 mM [-], respectively), as well as after PARP-1 silencing with PARP-1 targeted siRNA (200 nM) (gray triangle), but not after treatment with a nontargeted siRNA (C). Each value represents the average value calculated for four repetitions with a calculated variation coefficient (standard deviation divided by the average value) below 10% (Experimental Procedures; n = 2). (Bottom) PARP-1 expression was suppressed 48 hr after insertion of PARP-1-targeted siRNA (200 nM) by a transfection reagent into cultured brain cortical neurons in serum-deprived medium. PARP-1 expression was not affected by nontargeting siRNA (Experimental Procedures; n = 5).

treated neurons and cardiomyocytes (Figure 6B and Figures S4 and S5). Interestingly, suppression of PARP-1 activity or interference with PARP-1 expression by targeted siRNA suppressed Elk1 phosphorylation, while the phosphorylation of ERKs persisted (Figure 6B and Figure S5). These results are in line with the increased phosphorylation of r-Elk1 by polyADP-ribosylated r-PARP-1 in a protein mixture containing r-pERK2 and ATP (Figure 6A). PARP-1 Activation Mediates Expression of the Elk1 Target Gene c-fos in Cortical Neurons Treated with Nerve Growth Factor Activation of the transcription factor Elk1 by ERKcatalyzed phosphorylation induces expression of the immediate early gene c-fos (Herdegen and Leah, 1998; Buchwalter et al., 2004). We examined the effect of polyADP-ribosylated PARP-1 on the expression of the Elk1-

target gene c-fos. Levels of c-fos protein in the treated neurons were assayed by immunolabeling after protein extraction (Figure 6B), and c-fos gene expression was assayed by q-PCR (Figure 6C). Assay of the time course of Elk1 phosphorylation and of c-fos expression in cortical neurons treated with NGF disclosed that c-fos gene expression was significantly increased within 30 min after treatment with NGF and decreased thereafter (Figures 6B and 6C). As expected, preventing the phosphorylation of ERKs by MEK inhibition suppressed the phosphorylation of Elk1 and the expression of c-fos in the treated neurons (Figures 6B and 6C). Interestingly, the induced expression of c-fos gene was also suppressed when the expression of PARP-1 was reduced by PARP-1-targeted siRNA or when PARP-1 activity was suppressed by PARP-1 inhibitors (Figures 6B and 6C). These findings indicate that activated PARP-1



plays a role in ERK-catalyzed Elk1 phosphorylation and activation. PolyADP-Ribosylation Mediates ERK-Induced Acetylation of Core Histones Phosphorylation of the transcription factor Elk1 activates the histone acetyl transferase (HAT) activity of CBP/p300 (CBP, CREB-binding protein) (Li et al., 2003; Buchwalter et al., 2004). We therefore examined whether PARP-1 activation affects acetylation of core histones. This was tested in isolated nuclei after insertion of r-pERK2, in cortical neurons treated with nerve growth factors, and in cardiomyocytes treated with Ang II (Figures 7A and 7B and Figure S5). Core histone H4 was intensively acetylated in cultured cortical neurons treated with brain-derived neurotrophic factor (BDNF) (Figure 7), and the acetylation of core histones H3 and H4 was significantly enhanced in cardiomyocytes treated with Ang II (Supplemental Data; Figure S5). Agents preventing either ERK phosphorylation or polyADP-ribosylation significantly suppressed the acetylation of core histone H4 in the treated neurons (Figure 7A) and acetylation of H3 and H4 in the treated cardiomyocytes (Figure 5S). Acetylation of these core histones was similarly suppressed by PARP-1-targeted siRNA (Figure S5). These findings indicate that both PARP-1 activation and ERK phosphorylation are required for acetylation of core histones in the treated cells. The effects of PARP-1 activation and ERK2 phosphorylation on acetylation of core histone H4 were also examined in isolated nuclei of brain cortical neurons after insertion of r-pERK2. The inserted r-pERK2 induced polyADPribosylation of nuclear proteins (see also Figure 2A) as well as Elk1 phosphorylation and acetylation of histone H4 (Figure 7B). However, when the inserted r-pERK2 was replaced by the inactive recombinant KA-pERK2 (lacking the kinase activity of pERK2; Figure 4A), only polyADPribosylation persisted (as observed earlier, Figure 4A). Phosphorylated Elk1 and acetylated H4 were barely detectable (Figure 7B), indicating that Elk1 phosphorylation and H4 acetylation could be attributed to the activity of r-pERK2. Moreover, pERK-induced Elk1 phosphorylation and H4 acetylation were both suppressed when polyADPribosylation was suppressed (Figure 7B), indicating that polyADP-ribosylation is required for pERK-induced Elk1 phosphorylation and acetylation of core histone H4.

Figure 7. PARP-1 Activation Mediates Acetylation of Core Histone H4 in Cortical Neurons Treated with Nerve Growth Factor (A) Acetylated core histone H4 (green) and pERKs (red) were immunodetected by confocal imaging in primary cultures of brain cortical neurons treated with BDNF (100 ng/ml, 7 min, 37 C) before or after treatment (30 min incubation) with PARP inhibitor (PJ-34, 20 mM) or MEK inhibitors U0126 (10 mM) or PD098059 (50 mM), as indicated. Merged labeling is yellow (n = 5).

(B) PolyADP-ribosylation mediated core histone H4 acetylation, induced by insertion of r-pERK2 into nuclei of permeabilized cortical neurons in primary culture. r-pERK2 or its inactive but phosphorylated mutant KA-pERK2 (indicated as pERK2 and KA-pERK2) was inserted into the nuclei before or after treatment with the PARP-1 inhibitor PJ-34 (30 min, 10 mM). Acetylated histone H4 (green), phosphorylated Elk1 (red), and polyADP-ribosylated proteins (PAR, green) were immunodetected by confocal imaging (n = 3). (C) PARP-1 activation in the ERK signaling cascade. PARP-1, activated by pERK2 and polyADP ribosylated in the presence of NAD (upper panel), enhances pERK2-catalyzed phosphorylation of transcription factor Elk1, promoting the HAT activity of CBP/p300 and the expression of Elk1-target genes (lower panel).



DISCUSSION The results of this study point to an interplay between the activation of PARP-1 and ERK2 signaling. This was reflected here in a previously undocumented mode of PARP-1 activation by direct interaction with pERK2 and an increase in ERK2-catalyzed phosphorylation of the transcription factor Elk1 by polyADP-ribosylated PARP-1, promoting the expression of the Elk1-target gene c-fos. The findings in cell-free systems (Figures 2–4), in brain cortical neurons treated with nerve growth factors (Figures 6 and 7 and Figures S1 and S4) and in stimulated cardiomyocytes (Figure S5), provide the first evidence for an alternative mode of PARP-1 activation, which is independent of binding to DNA (Figures 3D and 4). Moreover, nicked DNA interfered with the activation of PARP-1 by pERK2 (Figure 3D). The DNA-independent activation of PARP-1 by pERK2 is consistent with a mechanism that depends on ERK2 phosphorylation, but not on its kinase activity, i.e., PARP-1 activation was not induced by phosphorylation (Figures 3B–3D, 4, and 7B; Supplemental Data). Thus, PARP-1 activation by pERK2 differs from phosphorylation-induced modulations in PARP-1 activated by nicked DNA (Kauppinen et al., 2006; Szabo et al., 2006). Several ERK2 substrates are activated by a direct interaction with phosphorylated ERK2 without involving its kinase activity (Camps et al., 1998; Shapiro et al., 1999). The properties of pERK2-activated PARP-1 differ from those of PARP-1 activated by nicked DNA. PARP-1 activated by pERK2 has a much higher affinity for NAD (Figures 4C and 4D); pERK2-activated r-PARP-1 was highly polyADP-ribosylated even at NAD concentration 50,000 times lower than the physiological concentration in the cytoplasm (Figures 3B, 3D, and 4C; Supplemental Data). Thus, NAD depletion caused by PARP-1 activation by damaged DNA (Szabo et al., 2006) is not anticipated for pERK-induced PARP-1 activation (Figure 4D and Supplemental Data). r-pERK2 bound with both r-PARP-1 and polyADP-ribosylated r-PARP-1 but preferentially with the latter (Figures 5C and 5D). This preference might result in a positive feedback mechanism (Figure 7C), which might keep PARP-1 polyADP-ribosylated as long as pERK2 is phosphorylated in the nucleus. Accordingly, the puzzling gradual decrease in pERK phosphorylation in the nucleus, despite the presence of phosphatases (Sasagawa et al., 2005), could prolong the active state of PARP-1 in cells stimulated by agents activating the ERK-signaling cascade. Also, polyADP-ribosylated PARP-1 could act as an anchoring protein for pERK2 in the nucleus (Lenormand et al., 1998). In a cell-free system, pERK2-induced PARP-1 polyADP-ribosylation was shown here to dramatically amplify the phosphorylation of Elk1 by pERK2 (Figure 6A). Accordingly, polyADP-ribosylation amplified Elk1 phosphorylation in cortical neurons and cardiomyocytes in response to signals activating the ERK cascade, and PARP-1 activation was required for the induced expression of the Elk1

target gene c-fos (Figures 6B and 6C). If we assume a direct association of Elk1 and PARP-1 with pERK2 (Figure 5D) in the nucleus, PARP-1 activation by pERK2 and the increase in Elk1 phosphorylation induced by polyADP-ribosylated PARP-1 might act as coupled events (Figure 7C), in which PARP-1 polyADP-ribosylated by pERK2 acts as a scaffold protein, increasing Elk1 phosphorylation by stabilizing the interaction of pERK2 with Elk1. Phosphorylation of the transcription factor Elk1 evokes the HAT activity of CBP/p300 (Li et al., 2003; Buchwalter et al., 2004). Thus, pERK2-induced PARP-1 activation (Figures 2–4 and 7B and Figure S5) and the increase in pERK-catalyzed Elk1 phosphorylation induced by polyADP-ribosylated PARP-1 (Figures 6A, 6B, and 7B and Figure S5) may underlie the enhanced acetylation of core histones in cortical neurons and cardiomyocytes treated with agents activating the ERK phosphorylation cascade (Figure 7A and Figure S5). Transcription factor Elk1 is activated by a variety of extracellular signals via the MAPK phosphorylation cascades (Buchwalter et al., 2004). In addition, expression of its target gene c-fos promotes the expression of other early genes (Herdegen and Leah, 1998; Sng et al., 2004). Thus, the present findings outline an epigenetic mechanism that enhances the expression of early genes in response to signals activating the ERK signaling cascade mediating growth and differentiation (Figure 7C). Accumulating findings indicate essential roles for secreted nerve growth factors, ERK2 and Elk1 phosphorylation, the HAT activity of CBP/p300, and expression of the immediate early gene c-fos in the formation of long-term memory (Impey et al., 1999; Thomson et al., 1999; Schinder and Poo, 2000; Kandel, 2001; Sng et al., 2004; Korzus et al., 2004; Martin and Sun, 2004). In addition, recent findings have shown that PARP-1 activation is required for long-term memory formation during learning (Cohen-Armon et al., 2004; A´. Fonta´n-Lozano et al., 2005, SFN Meeting 67.16, abstract). The enhanced expression of immediate-early genes by pERK2-induced PARP-1 activation might underlie the pivotal role of PARP-1 activation in memory consolidation. In view of the major role of gene expression regulated by the ERK phosphorylation cascade in cell proliferation, growth, and differentiation (Herdegen and Leah, 1998; Buchwalter et al., 2004), PARP-1 activation by phosphorylated ERK2 is expected to fulfill a variety of functions in the cell in the absence of damaged DNA.

EXPERIMENTAL PROCEDURES Cell Cultures, Protein Extracts, and Brain Slices Primary cultures were prepared from brain cortices of 18- to 19-dayold Sprague-Dawley rat embryos, as previously described (Homburg et al., 2000). Primary cultures of cardiomyocytes were prepared from ventricles of 1- to 2-day-old Sprague-Dawley rats (see Supplemental Data). Crude nuclei and nuclear protein extracts were prepared from cultured cells or from brain cortical slices, as described (Homburg et al., 2000; Visochek et al., 2005).



Recombinant ERK2 Recombinant ERK2 (r-ERK2) constructs were prepared as described (Jaaro et al., 1997). PolyADP-Ribosylation and [32P]PolyADP-Ribosylation of Purified or Recombinant PARP-1 PARP-1 prepared from nuclear extracts of calf thymus (40%–60% pure; BIOMOL) and human recombinant PARP-1 (r-PARP-1; >97% pure; Alexis) were polyADP-ribosylated by incubation (100 ng/sample; 20 min, 30 C) with b-NAD (0.2 mM; Sigma) and either sheared DNA (1 mg/sample) or r-pERK2 (100 ng/sample) in the presence of phosphatase and protease inhibitors. [32P]polyADP-ribosylation of PARP-1 in isolated nuclei was carried out as described (Homburg et al., 2000). For [32P]polyADP-ribosylation of r-PARP-1, r-PARP-1 (50–100 ng/ sample) was incubated with r-pERK2 (50–100 ng/sample) for 5–20 min at 30 C in 50 ml samples containing 10 mM MgCl2, 1.5 mM EGTA (pH 7.5), 1 mM DTT, phosphatase inhibitors (1 mM sodium orthovanadate, 25 mM b-glycerolphosphate), protease inhibitors (Sigma), and [32P]NAD (1000 Ci/mmol; 1 mCi/sample). The reaction was terminated by protein denaturation (addition of sample buffer and boiling for 1–2 min). Denatured proteins were mounted on SDSpolyacrylamide gels, transferred to nitrocellulose membranes (western blot), autoradiographed, and immunolabeled.

Exogenous DNA and Detection of DNA Breaks We used salmon sperm DNA (sheared double-stranded DNA; Sigma), DNA treated with DNase I (nicked DNA; Amersham), and intact genomic double-stranded DNA that was prepared from rat brains using the QIAamp DNA Mini Kit (Qiagen). Double-stranded DNA was examined for nicks formation by using alkaline gel electrophoresis (Homburg et al., 2000). 32

[ P]Phosphorylation by r-pERK2 [32P]Phosphorylation by r-pERK2 was carried out as described (Seger et al., 1992).

Two-Dimensional Gel Electrophoresis Two-dimensional gel electrophoresis was performed as described (Cohen-Armon et al., 2004; Visochek et al., 2005; Supplemental Experimental Procedures).

Insertion of r-pERK2 For insertion of r-pERK2 into permeabilized cells or isolated nuclei, we followed a previously described procedure (Matsubayashi et al., 2001; Supplemental Experimental Procedures).

Dot Blots The dot blot technique was used to detect interactions between recombinant proteins and protein binding to polyADP-ribosylated PARP-1 when coimmunoprecipitation was not possible (details included in Supplemental Experimental Procedures).

PARP-1 Silencing by Targeted siRNA The sequences 800–807 and 890–897 in the PARP-1 catalytic domain near the conserved active site (50 -AAGAUAGAGCGUGAAGGCGAA-30 and 50 -AAGCCUCCGCUCCUGAACAAU-30 ) were used to silence the expression of PARP-1. These are two of the five sequences tested before (Kameoka et al., 2004). A transfection reagent, siIMPORTER (Upstate Biotechnology) or X-tremeGENE siRNA Transfection Reagent (Roche Applied Science, Germany), was used to insert siRNA (200 nM; Dharmacon) into cultured cells in serum-deprived medium. PARP-1 expression was tested after 48 hr. As a control, we used a nontargeting siRNA (siRNA #2; Dharmacon).

Q-PCR Profiling of Relative Expression of the c-fos Gene in NGF-Treated Rat Brain Cortical Neurons The time course of the induced c-fos expression after treatment with NGF was estimated from the relative expression of segment 1734– 1853 in the gene by using 50 -GTTCCTGGCAATAGTGTGTTC-30 and 50 -GCTGAAGAGCTACAGTACGTG-30 as forward and backward primers, respectively, and quantitative PCR (see Supplemental Experimental Procedures for details). Supplemental Data Supplemental Data include Supplemental Results, Supplemental Experimental Procedures, five figures, and Supplemental References and can be found with this article online at http://www.molecule.org/ cgi/content/full/25/2/297/DC1/. ACKNOWLEDGMENTS This work was supported by The Israel Science Foundation, The National Institute for Psychobiology in Israel, The Kurt-Lion Foundation, The Israel Ministry of Health (M.C.-A.), and The Yale S. Lewine and Ella Miller Lewine Chair for Cancer Research (R.S.). Received: February 12, 2006 Revised: June 13, 2006 Accepted: December 12, 2006 Published: January 25, 2007 REFERENCES Buchwalter, G., Gross, C., and Wasylyk, B. (2004). Ets ternary complex transcription factors. Gene 324, 1–14. Buki, K.G., Bauer, P.I., Hakam, A., and Kun, E. (1995). Identification of domains of poly(ADP-ribose)polymerase for protein binding and self association. J. Biol. Chem. 270, 3370–3377. Camps, M., Nichols, A., Gillieron, C., Antonsson, B., Muda, M., Chabert, C., Boschert, U., and Arkinstall, S. (1998). Catalytic activation of the phosphatase MKP-3 by ERK2 mitogenic-activated protein kinase. Science 280, 1262–1264. Cohen-Armon, M., Visochek, L., Katzoff, A., Levitan, D., Susswein, A.J., Klein, R., Valbrun, M., and Schwartz, J.H. (2004). Long-term memory requires polyADP-ribosylation. Science 304, 1820–1822. Eblen, S.T., Catling, A.D., Assanah, M.C., and Weber, M.J. (2001). Biochemical and biological functions of the N-terminal, noncatalytic domain of extracellular signal-regulated kinase 2. Mol. Cell. Biol. 21, 249–259. Herdegen, T., and Leah, J.D. (1998). Inducible and constitutive transcription factors in the mammalian nervous system: control of gene expression by Jun, Fos and Krox and CREB/ATF proteins. Brain Res. Brain Res. Rev. 28, 370–490. Homburg, S., Visochek, L., Moran, N., Dantzer, F., Priel, E., Asculai, E., Schwartz, D., Rotter, V., Dekel, N., and Cohen-Armon, M. (2000). A fast signal-induced activation of poly(ADP-ribose) polymerase: a novel downstream target of phospholipase C. J. Cell Biol. 150, 293–308. Huang, E.J., and Reichardt, L.F. (2001). Neurotrophins: roles in neuronal development and function. Annu. Rev. Neurosci. 24, 295–318. Impey, S., Obrietan, K., and Storm, D.R. (1999). Making new connections: role of ERK-MAP kinase signaling in neuronal plasticity. Neuron 23, 11–14. Jaaro, H., Rubinfeld, H., Hanoch, T., and Seger, R. (1997). Nuclear translocation of mitogen-activated protein kinase kinase (MEK) in response to mitogenic stimulation. Proc. Natl. Acad. Sci. USA 94, 3742–3747. Kameoka, M., Nukuzuma, S., Itaya, A., Tanaka, Y., Ota, K., Ikuta, K., and Yoshihara, K. (2004). RNA interference directed against



polyADP-ribose polymerase I efficiently suppresses human immunodeficiency virus type 1 replication in human cells. J. Virol. 78, 8931– 8934.

Sasagawa, S., Ozaki, Y.-I., Fujita, K., and Kuroda, S. (2005). Prediction and validation of the distinct dynamics of transient and sustained ERK activation. Nat. Cell Biol. 7, 365–373.

Kandel, E.R. (2001). The molecular biology of memory storage: a dialogue between genes and synapses. Science 294, 1030–1038.

Schinder, A.F., and Poo, M. (2000). The neurotrophin hypothesis for synaptic plasticity. Trends Neurosci. 23, 639–645.

Kauppinen, T.M., Chan, W.Y., Suh, S.W., Wiggins, A.K., Huang, E.J., and Swanson, R.A. (2006). Direct phosphorylation and regulation of poly(ADP-ribose) polymerase-1 by extracellular signal regulated kinases 1/2. Proc. Natl. Acad. Sci. USA 103, 7136–7141.

Schreiber, V., Datzer, F., Ame¨, J.-C., and de Murcia, G. (2006). Novel functions for an old molecule. Nat. Rev. Mol. Cell Biol. 7, 517–528.

Korzus, E., Rosenfeld, M.G., and Mayford, M. (2004). CBP histone acetyltransferase activity is a critical component of memory consolidation. Neuron 42, 961–972.

Seger, R., Ahn, N.G., Posada, J., Munar, E.S., Jensen, A.M., Cooper, J.A., Cobb, M.H., and Krebs, E.G. (1992). Purification and characterization of mitogen-activated protein kinase activator MEK from epidermal growth factor-stimulated A431 cells. J. Biol. Chem. 267, 14373– 14381.

Kraus, W.L., and Lis, J.T. (2003). PARP goes transcription. Cell 113, 677–683. Lautier, D., Lagueux, J., Thibodeau, J., Menard, L., and Poirier, G.G. (1993). Molecular and biochemical features of poly(ADP-ribose) metabolism. Mol. Cell. Biochem. 122, 171–193. Lenormand, P., Brondello, J.-M., Brunet, A., and Pouyssegur, J. (1998). Growth factor-induced p42/p44 MAPK nuclear translocation and retention requires both MAPK activation and neosynthesis of nuclear anchoring proteins. J. Cell Biol. 142, 625–633.

Seger, R., and Krebs, E.G. (1995). The MAPK signaling cascade. FASEB J. 9, 726–735.

Shapiro, P.S., Whalen, A.M., Tolwinski, N.S., Wilsbacher, J., FroelichAmmon, S.J., Garcia, M., Osheroff, N., and Ahn, N.G. (1999). Extracellular signal regulated kinase activates topoisomerase-IIa through a mechanism independent of phosphorylation. Mol. Cell. Biol. 19, 3551–3560. Shaul, Y.D., and Seger, R. (2004). Use of inhibitors in the study of MAPK signaling. Methods Mol. Biol. 250, 113–126.

Li, Q.J., Yang, S.H., Maeda, Y., Sladek, F.M., Sharrocks, A.D., and Martins-Green, M. (2003). MAP kinase phosphorylation-dependent activation of Elk-1 leads to activation of the co-activator p300. EMBO J. 15, 281–291.

Sng, J.C.G., Taniura, H., and Yoneda, Y. (2004). A tale of early response genes. Biol. Pharm. Bull. 27, 606–612.

Martin, K.C., and Sun, Y.E. (2004). To learn better, keep the HAT on. Neuron 42, 879–881.

Thomson, S., Mahadevan, L.C., and Clayton, A.L. (1999). MAP kinase mediated signaling to nucleosomes and immediate-early gene induction. Semin. Cell Dev. Biol. 10, 205–214.

Szabo, C., Pacher, P., and Swanson, R.A. (2006). Novel modulators of poly(ADP-ribose) polymerase. Trends Pharmacol. Sci. 27, 626–630.

Matsubayashi, Y., Fukuda, M., and Nishida, E. (2001). Evidence for the existence of a nuclear pore complex-mediated, cytosol-independent pathway of nuclear translocation of ERK MAP kinase in permeabilized cells. J. Biol. Chem. 276, 41755–41760.

Tulin, A., and Spradling, A. (2003). Chromatin loosening by poly(ADP)ribose polymerase (PARP) at Drosophila puff loci. Science 299, 560– 562.

Mendoza-Alvarez, H., and Alvarez-Gonzalez, R. (1993). PolyADPribose polymerase is a catalytic dimer and the automodification reaction is intramolecular. J. Biol. Chem. 268, 22575–22580.

Visochek, L., Vulih, I., Steingart, R.A., Klein, R., Priel, E., Gozes, I., and Cohen-Armon, M. (2005). PolyADP-ribosylation is involved in neurotrophic activity. J. Neurosci. 25, 7420–7428.

Menissier-de Murcia, J., Molinete, M., Gradwohl, G., Simonin, F., and de Murcia, G. (1989). Zinc-binding domain of poly(ADP-ribose)polymerase participates in the recognition of single strand breaks on DNA. J. Mol. Biol. 210, 229–239.

Wolf, I., Rubinfeld, H., Yoon, S., Marmor, G., Hanoch, T., and Seger, R. (2001). Involvement of the activation loop of ERK in the detachment from cytosolic anchoring. J. Biol. Chem. 276, 24490–24497.

Pouyssegur, J., and Lenormand, P. (2003). Fidelity and spatio-temporal control in MAP kinase (ERKs) signaling. Eur. J. Biochem. 270, 3291– 3299. Rouleau, M., Aubin, R.A., and Poirier, G.G. (2004). PolyADP-ribosylated chromatin domains: access granted. J. Cell Sci. 117, 815–825.

Yamazaki, T., Komuro, I., and Yazaki, Y. (1999). The molecular mechanism of cardiac hypertrophy and failure. Ann. N Y Acad. Sci. 874, 38–48. Yoon, S., and Seger, R. (2006). The extracellular signal-regulated kinase: Multiple substrates regulate diverse cellular functions. Growth Factors 24, 21–44.
