|Home | About | Journals | Submit | Contact Us | Français|
Apurinic/apyrimidinic endonuclease 1 (APE1), an essential protein in mammals, is involved in base excision DNA repair (BER) and in regulation of gene expression, acting as a redox co-activator of several transcription factors. Recent findings highlight a novel role for APE1 in RNA metabolism, which is modulated by nucleophosmin (NPM1). The results reported in this article show that five lysine residues (K24, K25, K27, K31 and K32), located in the APE1 N-terminal unstructured domain, are involved in the interaction of APE1 with both RNA and NPM1, thus supporting a competitive binding mechanism. Data from kinetic experiments demonstrate that the APE1 N-terminal domain also serves as a device for fine regulation of protein catalytic activity on abasic DNA. Interestingly, some of these critical lysine residues undergo acetylation in vivo. These results suggest that protein–protein interactions and/or post-translational modifications involving APE1 N-terminal domain may play important in vivo roles, in better coordinating and fine-tuning protein BER activity and function on RNA metabolism.
APE1/Ref-1 (APE1) is an essential protein acting as a master regulator of cellular responses to oxidative stress and contributing to the maintenance of genome stability (1–3). APE1 is involved both in the base excision repair (BER) of DNA lesions, functioning as the major DNA AP-endonuclease in mammals, and in transcription, acting as a redox co-activator of different transcription factors such as Egr-1, NF-κB and p53 (4,5). These biological activities are located in two functionally distinct protein regions. The N-terminal domain (residues 1–127), containing the nuclear localization signal (NLS), is mainly devoted to the transcriptional co-activating function, while the C-terminal portion (residues 61–318) exerts the enzymatic activity on abasic sites in DNA (4–7). A third function, regulated by K6/K7 acetylation (8), is its transcriptional repressor activity through indirect binding to the negative Ca2+-response elements (nCaRE) (9,10). A completely unexpected new function has been recently unveiled and is associated to a role in RNA metabolism (11–13), which could explain some of the effects of APE1 on gene expression through post-transcriptional mechanisms.
Extensive structural investigations have shown that the first 42 amino acids of APE1 are disordered, while the remainder of the protein has a globular fold (14–16). APE1 homology with Escherichia coli ExoIII starts from residue 62; a Blast homology search against non-redundant databases clearly shows that the protein C-terminus is highly conserved, while the N-terminus (first ~40 residues) is not. Conservation of APE1 N-terminal portion is high in mammals (>90%) but almost absent in other organisms, with the exception of Danio, Drosophila, Xenopus and Dictyostelium (having a sequence identity <40%), suggesting that it may be a recent acquisition during evolution (17). Interestingly, the Drosophila homologue of mammalian APE1 (i.e. Rrp1), which also bears a bipartite arrangement, presents an unstructured N-terminal domain which is used to interact with Pol ζ (18,19). Accordingly, data on the human APE1 (hAPE1) clearly show that protein N-terminus (first ~35 amino acids) is used for interacting with other partners, such as XRCC1 (20), CSB (21), NPM1 and others (11), possibly modulating the different APE1 functions. These data underscore a common emerging theme in this protein: the presence of an unstructured N-terminus for mediating protein–protein interaction. Its truncation has been related to an enhanced immune cell death mediated by granzyme A and K (22,23), to a reduction of hAPE1 accumulation within nuclei (22,24) and to a loss of its interaction with XRCC1 (20) and other protein partners (11). A truncated form has also been detected in mitochondria, where it could exert a role in mtDNA repair (25–27). As the protein lacking the first 33 residues (hAPE1NΔ33) has been associated with an apoptotic phenotype (23), it cannot be excluded that its generation may promote a cytotoxic outcome, driving pro-apoptotic triggering directly within mitochondria.
hAPE1 is mainly a nuclear protein and is critical for controlling cellular proliferative rates (1,2,28). However, cytoplasmic, mitochondrial and ER localizations have also been reported (25–27,29,30). hAPE1 is an abundant and relatively stable protein in mammalian cells (4,5). Fine-tuning of the multiple hAPE1 functions may therefore depend on post-translational modifications (PTMs) and on its interactome modulation. While a functional role has been determined for some PTMs (K6/K7 acetylation and K24/K25/K27 ubiquitination) (8,31,32), the identity and the role of the interacting partners in modulating hAPE1 biological functions are still under investigation (11,31,33). By a proteomic approach, we recently discovered that hAPE1, through its 33N-terminal amino acids (33N-term), may functionally interact with different proteins involved in RNA metabolism, unveiling its unexpected function in RNA cleansing processes (11,12). We demonstrated that hAPE1 may affect cell growth by directly acting on RNA quality-control mechanisms, thus regulating gene expression through post-transcriptional mechanisms; in particular, hAPE1 interaction with NPM1 may affect its activity over rRNA molecules. However, many aspects of this new function are still unknown (12).
In this study, we explore the role of hAPE1 N-terminal domain by analysing the molecular details of its binding to both RNA and NPM1 and their impact on the DNA repair activity of the protein. Five lysine residues, located in the unstructured hAPE1 N-terminal domain made of 33 amino acids (33N-term) probably acquired during evolution, resulted crucial for hAPE1 interaction with RNA and NPM1. Interestingly, we found that some of these lysine residues resulted acetylated in vivo. Moreover, kinetic/functional studies showed that the unstructured 33N-term serves as a device for subtle regulation of hAPE1 catalytic activity on abasic DNA through regulation of product binding. These findings suggest that other proteins can modulate hAPE1 BER activity and ensure the correct processing of DNA intermediates and may have an impact on the RNA cleansing process. These data shed new light on the fine-tuning mechanisms responsible for hAPE1 regulation and open the way for developing new strategies for therapeutical intervention based on the modulation of hAPE1 functions.
Biophysical parameters of different APE1 peptides (molecular weight and pI) were calculated by the computing algorithms available at the Swiss Institute of Bioinformatics (http://www.expasy.ch/tools/pi_tool.html).
For phylogenetic analysis, a total of 17 APE1 orthologous metazoan sequences (Homo sapiens, Pan troglodytes, Macaca mulatta, Equus caballus, Sus scrofa, Bos taurus, Canis lupus familiaris, Mus musculus, Rattus norvegicus, Ornithorhyncus anatinus, Xenopus laevis, Xenopus tropicalis, Salmo salar, Danio rerio, Gasterosteus aculeatus, Caenorhabditis elegans and Strongylocentrotus purpuratus) were retrieved from databases using the BLAST algorithm (34). Out-group sequences were not included because of the almost complete lack of the N-terminal domain in other organisms. Alignment of the amino acid residues 1–60 and of the cDNA nucleotide 1–180 was made with the MAFFT program (v6.531b) (35) with default parameters, and evolutionary trees were inferred using the software MEGA version 4 (36). The minimum evolution tree-building method was used to reconstruct the phylogenetic tree (37); dendrograms’ topologies were checked by the bootstrap method with 1000 replicates.
HeLa cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% foetal calf serum (FCS), 100U/ml penicillin and 10μg/ml streptomycin sulphate. One day before transfection, cells were seeded in 10cm plates at a density of 1.8×106cells/plate. Cells were then transiently transfected with 2µg of either pFLAG5.1 (empty vector) or pFLAG5.1-hAPE1 (wild-type and specific mutant forms) plasmids per dish, using Lipofectamine 2000 Reagent (Invitrogen, Milan, Italy) according to the manufacturer’s instructions. Cells were harvested 48h after transfection. Co-immunoprecipitation analysis was performed as described earlier (11).
The protein samples were electrophoresed onto a 10–12% sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS–PAGE). Proteins were then transferred to nitrocellulose membranes (Schleicher & Schuell, Keene, NH, USA), revealed as already described (11) using ECL chemiluminescence procedure (GE Healthcare, Milan, Italy) and quantified by using a Gel Doc 2000 video-densitometer (Bio-Rad, Milan, Italy).
The constructs pGex3X-hAPE1 and pGex-3X-zAPE1, coding for the glutathione S-transferase (GST)-hAPE1 and zebrafish zAPE1 full-length proteins, respectively, were kindly provided by Mark R. Kelley (Indiana University, IN, USA). The constructs pGex3X-NΔ33hAPE1, pGex3X-Δ22hAPE1 and pGex3X-Δ9hAPE1, coding for truncated hAPE1 forms were generated by subcloning from the previous one through PCR and confirmed by direct sequencing. Plasmids pGex3X-hAPE1K3 mutant (K24A, K25A, K27A) and pGex3X-hAPE1K5 mutant (K24A, K25A, K27A, K31A, K32A) were generated using the Quickchange II XL Mutagenesis kit (Stratagene, Cedar Creek, TX, USA) and the pGex3X-hAPE1 as template, following manufacturer’s instructions. The plasmid coding for GST-NPM1 full-length was kindly provided by P. G. Pelicci (European Institute of Oncology, Milan, Italy). For expression of His6-tagged NPM1, cDNA coding for full-length NPM1 was subcloned in the pET-15b-expressing plasmid (Invitrogen).
All proteins were expressed in E. coli BL21(DE3), induced with 1mM isopropyl-β-d-thiogalactopyranoside (IPTG), and then purified on an AKTA Prime FPLC system (GE Healthcare) by using a GSTrap HP column (GE Healthcare) (for GST-tagged proteins) or a HisTrap HP column (GE Healthcare) (for His6-NPM1). The quality of purification was checked by silver-stained SDS–PAGE analysis. Extensive dialysis against PBS was performed to remove any trace of imidazole from the HisTrap-purified proteins.
Accurate quantification of all recombinant proteins was performed by colorimetric Bradford assays (Bio-Rad) and confirmed by SDS–PAGE and western blotting analysis using anti-GST peroxidase-conjugated antibody (Sigma, Milan, Italy). To remove the GST tag, GST–hAPE1 and GST–hAPE1 mutant proteins were further hydrolyzed with factor Xa (five factor Xa units per milligram of recombinant GST-fused protein). The protease was removed from the sample using a benzamidine HiTrap FF column (GE Healthcare); the proteins were then purified on a GSTrap HP column (GE Healthcare).
The correct molecular mass of each recombinant protein was assessed by LC–MS analysis.
Reagents for peptide synthesis were from Novabiochem (Laufelfingen, Switzerland) and Inbios (Pozzuoli, Italy). Solid phase peptide syntheses were performed on an automated synthesizer Syro I (MultiSynTech, Witten, Germany). Preparative RP–HPLC were carried out on a Shimadzu LC-8A, using a Phenomenex Luna-COMBI HTS C18 column (5×2.12cm, 10µm). LC–ESI–Ion Trap (IT)-MS analysis was carried out on a Surveyor HPLC system connected to a LCQ DECA XP IT (Thermo Scientific, USA), hAPE1 peptide 1–33 was purchased from EzBiolab (Westfield, IN, USA) and used without any further purification. Peptides corresponding to hAPE1 fragments 1–9 (in a free and acetylated version in K6 and K7) and 23–33 were prepared following standard Fmoc (9-fluorenylmethoxycarbonyl) chemistry protocols (38,39). Products were purified to homogeneity by RP–HPLC and their identity and purity were confirmed by LC–MS analysis.
An amount of 0.12nmol of either GST–hAPE1 or each GST–hAPE1 mutant protein were added, together with 0.2nmol of NPM1 protein or 10µg of total RNA, to 10µl of glutathione–Sepharose 4B beads (GE Healthcare). RNA was extracted using Trizol® Reagent (Invitrogen). Binding was performed in PBS supplemented with 5mM DTT and 1×protease inhibitor cocktail (Sigma) for 2h, under rotation, at 4°C. The beads were washed three times with washing buffer (PBS) supplemented with 0.1% w/v Igepal CA-630 (Sigma), 5mM DTT, 1×protease inhibitor cocktail. Beads were then resuspended in 1×Laemmli sample buffer containing 100mM DTT for western blot analysis or in Trizol® reagent for RNA extraction and rRNA quantification. For western blot experiments, GST-tagged and NPM1 proteins were detected with an anti-GST peroxidase-conjugated antibody (Sigma) and an anti-NPM1 monoclonal antibody (NPM1: clone FC-61991; Invitrogen), respectively.
Co-immunoprecipitation analyses of hAPE1 and its different mutants with NPM1 were performed as described earlier (11).
Levels of 28S and 18S rRNA transcripts were determined as described earlier (11) by using the iScript cDNA synthesis kit (Bio-Rad) according to the manufacturer's instructions. Q-PCR was performed with an iQ5 multicolor real-time PCR detection system (Bio-Rad), according to the manufacturer’s protocol. The following primers were used: 18S_for (5′-CTGCCCTATCAACTTTCGATGGTAG-3′) and 18S_rev (5′-CCGTTTCTCAGGCTCCCTCTC-3′), which amplified a region of 100bp; 28S_for (5′-TGTCGGCTCTTCCTATCATTGT-3′) and 28S_rev (5′-ACCCAGCTCACGTTCCCTATTA-3′), which amplified a region of 81bp.
Each sample analysis was performed in triplicate. As negative control, a sample without the template was used; as control for genomic DNA contamination, a sample with non-retro-transcribed RNA was included instead of template cDNA. The cycling parameters were denaturation at 95°C for 10s and annealing/extension at 60°C for 30s (repeated 40 times). In order to verify the specificity of the amplification, a melting-curve analysis was performed, immediately after the amplification protocol. A standard curve was generated by using a calibrating cDNA. This cDNA was obtained from a retrotranscription of 1µg of total RNA from HeLa cells, which was serially diluted and analysed for 18S and 28S rRNA. We assumed that rRNA represented 90% of the total RNA, in a 28S/18S rRNA molar ratio of 2:1. By using the equation describing the plots of the log10 of the starting amount of five dilutions of the calibrator cDNA versus the corresponding threshold cycle, the iQ5 optical system software calculated the amount of the template for each sample.
For experiments on hAPE1-Flag/NPM1 colocalization, a double immunofluorescence procedure was carried out as described earlier (11), where hAPE1 was labelled by an anti-Flag antibody-FITC conjugate (Sigma), whereas NPM1 was labelled with the Zenon® Mouse IgG Labeling Kit (Invitrogen), using the Alexa Fluor 546-labelled Fab fragment directed against the Fc portion of the IgG primary anti-NPM1 antibody (Invitrogen), according to the manufacturer’s instructions. Nuclei were counterstained with DAPI (Sigma); microscope slides were mounted and visualized through a Leica TCS SP laser-scanning confocal microscope.
Cleavage assays were performed as described earlier (11,20) with some minor modifications and using a fluorescent abasic dsDNA probe called 34FDNA (11). When the effect of other proteins on the hAPE1 activity was investigated, proteins were first incubated in 5mM Tris–HCl, pH 7.6, 5mM KCl, 1mM MgCl2, 0.1mg/ml bovine serum albumin and 0.005% Triton X-100, on ice, for 20min. Then, 150fmol of the labelled duplex containing a THF residue were added to each sample (11). Reactions were carried out at 37°C, for 30min, and stopped by adding 4µl of formamide dye (10mM EDTA, 0.5% Bromophenol Blue, 80% formamide) followed by heating for 5min, at 95°C, before loading. Reaction products were separated by denaturing 20% PAGE (19:1 acrylamide:bis-acrylamide). Gels were scanned and band intensities were quantified using a Storm Phosphoimager (Molecular Dynamics).
hAPE1–RNA complex formation was assessed as described earlier with a 34-mer synthetic RNA oligonucleotide containing an abasic site called 34FRNA (11), incubated on ice for 10min and separated onto a native 8% w/v polyacrylamide gel at 120V, for 2h. Each reaction contained 2.5pmol of 34FRNA labelled substrate and the indicated amount of proteins in a final volume of 10µl.
Real time binding assays were performed on a Biacore 3000 Surface Plasmon Resonance (SPR) instrument (GE Healthcare). Full-length hAPE1 and hAPE1NΔ33 were immobilized on a CM5 Biacore sensor chip in 10mM sodium acetate, pH 5.5, by using the EDC/NHS chemistry, with a reaction flow rate of 5μl/min and a time injection of 7min (40). Binding assays were carried out by injecting 90µl of analyte, at 20µl/min, with HBS (10mM HEPES, 150mM NaCl, 3mM EDTA, pH 7.4), 0.1mM TCEP as running-buffer. BIAevaluation analysis package (version 4.1, GE Healthcare), was used to subtract the signal of the reference channel, while RUmax data versus proteins concentrations were fitted with GraphPad Prism, vers. 4.00 (San Diego, CA), following a procedure reported earlier (41).
Identification of PTMs occurring on hAPE1 was performed on immunopurified hAPE1 obtained from hAPE1-Flag expressing HeLa cells (11). hAPE1 was resolved by 12% SDS–PAGE; corresponding protein band was excised, in-gel S-alkylated, digested with endoprotease AspN and analysed as reported earlier (11,42). Details on mass spectrometry analysis are described in the Supplementary Data.
Statistical analysis was performed using the Microsoft Excel data analysis program for Student’s t-test analysis. P<0.05 was considered as statistically significant.
NPM1 and RNA interact with hAPE1 through its 33N-term (11). In order to map the specific residues responsible for binding, we performed a preliminary analysis of the 33N-term sequence. Based on the distribution of charged amino acids at physiological pH, the 33N-term can be subdivided into three distinct regions: (i) a first basic sequence comprising amino acids 1–9, (ii) an acidic sequence comprising the amino acids 10–22 and (iii) a second basic region, spanning amino acids 23–33 and containing five lysine residues (Figure 1A). Interestingly, the 33N-term confers a calculated basic pI (8.3) to the full-length protein, while its removal decreases the pI to 6.6. These calculations were supported by experimental evaluation of the pI of the two protein forms through 2-DE analysis (Supplementary Figure S1).
Previous in vitro experiments demonstrated that hAPE1 is able to bind RNA (11). In this work, we verified that this process reasonably occurs also in vivo, as demonstrated by co-localization analysis of hAPE1 with RNA in HeLa cells (Supplementary Figure S2). Due to its basic features, the 33N-term peptide could potentially bind to RNA by itself. We therefore tested peptide binding to RNA by EMSA, as described earlier (11). Figure 1B (left) shows that the peptide 1–33 binds a 34-mer RNA oligonucleotide containing an abasic sites (34FRNA) (11) and competes for its binding with the full-length hAPE1 protein. More striking results were obtained with rRNA, through GST pull-down competition experiments (Figure 1B, right) and with t-RNA molecules of different origins (data not shown).
We mapped the amino acids responsible for RNA binding through deletion analyses and competition assays with peptides spanning the regions 1–9, 23–33 and 1–33. The binding abilities of hAPE1NΔ9, hAPE1NΔ22 and hAPE1NΔ33 (Figure 2A) were evaluated through GST pull-down experiments with total RNA samples purified from HeLa cells as prey (11). Figure 2B shows that amino acid residues 23–33 are required for efficient rRNA binding by hAPE1, as also confirmed by competition experiments with the specific 1–9, 23–33 and 1–33 peptides (Figure 2C). Acetylation of K6 and K7 residues, which regulates hAPE1 binding to nCaRE sequences (8), does not influence the hAPE1 RNA-binding activity, as determined by competition experiments with the acetylated peptide, i.e. 1-9ac hAPE1 (Figure 2C). The reported experiments, relative to hAPE1 binding to 28S rRNA, were also performed with 18S rRNA giving similar results (not shown).
Overall, these experiments demonstrate a leading role for the 23–33 basic amino acid sequence in mediating hAPE1 binding to RNA.
Since electrostatics may drive the RNA-binding activity of hAPE1, we tested the hypothesis that basic residues within the 23–33 hAPE1 sequence are responsible for mediating stable binding to RNA. Thus, we generated mutants of the protein bearing a triple (K24/K25/K27) and quintuple (K24/K25/K27/K31/K32) K to A substitutions, here referred to as hAPE1K3 and hAPE1K5, respectively. Recombinant proteins were expressed in E. coli and chromatographycally purified. SDS–PAGE analyses of purified proteins showed that each mutant displays an altered mobility (Figure 3A, left), possibly due to alteration of its overall charge since LC–MS analyses confirmed the correctness of their masses and the difference in their apparent mobility observed in SDS–PAGE was abolished when separating proteins in urea-containing denaturing gels (not shown). Then, their binding activity to 34FRNA was measured through EMSA analysis (Figure 3A, right). The obtained results clearly demonstrate that the hAPE1 binding to this small RNA strictly depends on the presence of these lysine residues. GST pull-down experiments with 28S and 18S rRNA (Figure 3B and data not shown) were in good agreement with the EMSA experiments. As expected, hAPE1K5 mutant was as deficient in binding to both 28S (Figure 3B) and 18S rRNA (not shown) as the hAPE1NΔ33 deletion mutant. On the contrary, the hAPE1K3, which was proficient in binding to 28S RNA, as shown in Figure 3B, showed only a slightly reduced ability to bind 18S RNA compared to the entire hAPE1 (not shown).
We (11) and others (25) previously reported that the enzymatic activity on abasic dsDNA (34FDNA) of the hAPE1NΔ33 protein was significantly higher than that of the wild-type protein. We therefore evaluated the effect of the K to A mutations on the endonuclease activity of APE1 over abasic dsDNA, by measuring the kinetic parameters (KM and kCAT) of hAPE1, hAPE1NΔ33 and the two K to A mutants (Table 1). Since the different proteins exhibited very different endonuclease activities, an initial set of experiments (Figure 4A) was performed in order to select the most suitable enzyme concentration for further analysis. An enzyme concentration capable of cleaving ~7nM 34FDNA over the time of reaction, indicated by the red spots, was selected for time-lapse experiments and its value is indicated in the plots. KM and kCAT values for each enzyme are shown in Table 1, and are in good agreement with the parameters reported earlier by other authors (25,43).
Our findings (Table 1) show that the KM value is similar for all protein variants, therefore suggesting that the enzyme affinity for the abasic site in the 34FDNA substrate is not significantly affected by the removal of the 33N-term or by mutation of the critical lysine residues within the region 23–33. On the contrary, kCAT values are very different, indicating that the enzymatic catalysis is faster for the NΔ33 or the K5 mutants than for the full-length protein. These experiments were performed with different preparations of recombinant proteins confirming the reliability of our data.
These findings suggest that molecular interactions and electrostatics involving the amino acid region 23–33 of hAPE1, though not appreciably changing the binding affinity to a double-stranded abasic substrate, participate in the modulation of hAPE1 endonuclease activity.
Such a model would predict a differential sensitivity of the endonuclease activity of the enzyme to ionic strength variation for the different hAPE1 mutants here investigated. We therefore studied the effect of K+ concentration on the catalytic activity of hAPE1 and of its mutants at concentrations of each enzyme capable of cleaving the same amount of 34FDNA (as performed for the experiments reported in Table 1). Data reported in Figure 4B clearly show that, as expected from previous results (43), the hAPE1 endonuclease activity was at least 4-fold higher in 150mM KCl than in 40mM KCl. Consistently with the proposed model, the catalytic activities of both K5 and NΔ33 protein mutants were essentially not affected by the increase in K+ concentration. As expected, the K3 mutant showed an intermediate behaviour. Accordingly, further enzymatic assays, performed as a function of salt concentration and at the same protein concentrations, show that while hAPE1 displays a maximum activity at 150mM KCl, hAPE1NΔ33 is more active at 80mM KCl as the K5 mutant is (Supplementary Figure S3).
Altogether, these data support a model in which the electrostatic charges of lysine residues in the region 23–33 are involved in the modulation of hAPE1 catalytic activity, disclosing new unsuspected details in its functional regulation. They also suggest that molecular interactions, or even PTMs, involving that region, act as strong modulators of hAPE1 endonuclease activity.
We reported earlier that hAPE1 interacts in vivo with NPM1 and that this interaction determines the localization of hAPE1 in nucleoli (11). To evaluate the dissociation constant (KD) of complexes formed between these proteins and their variants, SPR experiments were performed. The overlay of relative sensorgrams is reported in Figure 5A, left. The value of KD of the hAPE1/NPM1 complex was estimated by plotting RUmax values, from each SPR-binding experiment, as a function of NPM1 concentration (Figure 5A, middle). Data fitting using a non-linear regression analysis provided a KD value of 3.5±0.3µM. Experiments performed under the same conditions on the immobilized hAPE1NΔ33 gave a poor signal variation (the overlay of the sensorgrams is reported in Figure 5A, right). The degree of mean RUmax and the lack of dose-response dependence confirmed the absence of a stable binding between hAPE1NΔ33 and NPM1, as reported earlier (11).
In line with previous findings, SPR experiments also indicated that the N-terminal domain of NPM1 (residues 1–117) is responsible for the binding to hAPE1. In this case, the calculated KD value was 0.668±0.005µM (Supplementary Figure S4). Collectively, these data provided support to the hypothesis that the NPM1–hAPE1 interaction is physiologically relevant and that the 1–33 region of hAPE1 is responsible for it (11).
Then, we mapped the hAPE1 amino acids involved in NPM1 interaction through GST pull-down experiments using the hAPE1NΔ9, hAPE1NΔ22 and hAPE1NΔ33 deletion mutants. Similarly to what was observed for RNA binding, hAPE1 amino acids within region 23–33 is critical for the interaction with NPM1 (Figure 5B). Notably, removal of the first 22 amino acids, including the acidic stretch 10–22, increased over 2-fold the hAPE1 affinity to NPM1, thus suggesting that electrostatics plays a role in the interaction between the two proteins. GST pull-down experiments performed with the K to A mutants pointed out that, similarly to what seen for RNA binding, the presence of K31 and K32, in addition to K24, K25 and K27, is required for stable binding of hAPE1 to NPM1 (Figure 5C).
To confirm our in vitro experiments in an in vivo transfection assay, the two hAPE1K3 and hAPE1K5 mutants were expressed as Flag-tagged proteins in HeLa cells. hAPE1NΔ33 was used as a control. Notably, the two mutant proteins (i.e. K3 and K5), particularly the K5, resulted expressed at a slightly lower levels than the wild-type, suggesting that this mutation may alter protein stability. The hAPE1K5 mutant showed a behaviour similar to that of hAPE1NΔ33 (11), with a predominant cytoplasmic localization and no colocalization with NPM1 (Figure 5D, left). The hAPE1K3 mutant, on the contrary, showed an intermediate phenotype (Figure 5D, left). Co-immunoprecipitation analysis (Figure 5D, right) of transiently transfected HeLa cells confirmed the co-immunofluorescence data showing no interaction between hAPE1K5 and NPM1 as that of hAPE1NΔ33. We can therefore infer that the NLS sequence is not sufficient to guarantee for a complete nuclear localization of hAPE1 in the absence of charged residues at position K24, K25, K27, K31 and K32, thus suggesting that the RNA- and/or the NPM1-binding abilities may play a significant role in hAPE1 nuclear maintenance, as hypothesized earlier (11).
Taken together, these data provide the molecular basis for explaining the inhibitory effect of NPM1 on hAPE1 RNA binding observed previously (11). To better substantiate this, we performed competition experiments with NPM1 and hAPE1 in RNA-binding assays by GST pull-down analysis. Figure 5E shows that hAPE1 directly binds rRNA molecules, but the addition of soluble recombinant NPM1 in a 50% molar excess to the GST–hAPE1 bait inhibits hAPE1 binding to rRNA. This effect seems due to protein–protein interaction between hAPE1 and NPM1, since the rRNA-binding activity of GST-NPM1 for the 28S rRNA was lower than that of GST–hAPE1 alone. These data again support the notion that protein–protein interactions account for the inhibitory effect of NPM1 over hAPE1 binding to RNA, as suggested earlier (11).
As NPM1 stimulates hAPE1 endonuclease activity on 34FDNA substrates (11), we investigated whether these critical lysine residues may play a role in enzyme activation. Endonuclease assays under standard conditions at 40mM KCl, either with or without NPM1, showed that this protein stimulates only the activity of the wild-type hAPE1 (Figure 5F) in a way similar to that described earlier (11). We then checked whether this stimulatory effect could be observed at higher K+ concentration (150mM KCl), where binding of hAPE1 to NPM1 still occurs (11). As expected, no effect was observed for any of the enzymes tested. Thus, these results support the hypothesis that the hAPE1 region 23–33, through electrostatics, behaves as a site of allosteric regulation of the enzyme endonuclease activity.
K to A mutation mimicks constitutive acetylation in vivo. To probe whether the critical lysines we studied in the present work are acetylated in vivo, Flag-tagged hAPE1 from stably expressing HeLa cells (11) was immunopurified, excised from SDS–PAGE, S-alkylated, digested with endoprotease AspN, and subjected to massive nanoLC–ESI–LIT–MS/MS and MALDI–TOF–MS experiments. The protease was chosen to generate peptides having suitable dimension/properties for optimal PTMs mapping analysis of the 15–35 amino acid region.
In addition to a number of non-modified peptides, nanoLC–ESI–LIT–MS/MS analysis revealed the occurrence of specific acetylated species bearing a variable modification extent. In particular, peptides (15–33)Ac, (15–33)Ac2, (15–33)Ac3, (15–35)Ac, (15–35)Ac3 and (15–35)Ac4 at m/z 2186.4, 2228.3, 2270.4, 2429.1, 2513.3 and 2555.5 (theoretical value: 2186.2, 2228.2, 2270.2, 2429.3, 2513.3 and 2555.3), respectively, were detected in the total ion current MS trace profile of the hAPE1 digest. Strong signals corresponding to the non-acetylated peptide counterparts were also observed (data not shown). Confirmation of the peptide assignment and direct localization of the acetylation sites was obtained from MS/MS data. As representative examples, the fragmentation spectra of peptides (15–33)Ac3 and (15–35)Ac4 are presented in Figure 6. Accordingly, nanoLC–ESI–LIT–MS/MS experiments demonstrated that hAPE1 is acetylated at K27, K31, K32 and K35. Based on the data from MALDI–TOF–MS analysis that, combined to nanoLC–ESI–LIT–MS/MS, allowed coverage of 68% of the protein sequence, no additional PTMs were observed within this peptide region. Furthermore, the reliability of our approach was substantiated by the analysis of the hAPE1 tryptic digest, which confirmed acetylation occurring at K6 and K7 (data not shown), as demonstrated earlier (8).
While being highly conserved in mammals, the APE1 33N-term is poorly conserved in other organisms including some vertebrates (5,17). We sought divergences in APE1 33N-terms within vertebrates in order to correlate them with protein binding to RNA and endonuclease activities. An evolutionary tree depicting the relationship between the vertebrate sequences is shown in Figure 7A. The sequences from amphibians and fishes represent a separate branch from mammalian ones. As reported earlier for the Cys content analysis, it turned out that, based on the N-terminal portion of the protein, zebrafish APE1 (zAPE1) is evolutionarily the most distant sequence from hAPE1 within vertebrate sequences (17). Alignment analysis (Figure 7B) suggests that during phylogenesis, the first three lysine residues (i.e. K24, K25 and K27) were conserved, while the other two (i.e. K31 and K32), occurring in mammals and Xenopus, might come from an insertion. Thus, fishes’ APE1, which lacks these lysine residues, should act as a loss-of-function protein with regard to its RNA-binding. To test this hypothesis, we analysed the RNA-binding activity of zAPE1 in comparison with that of hAPE1. EMSA (Figure 7C, left), performed with purified recombinant proteins (Supplementary Figure S5), clearly demonstrated that zAPE1 binds to RNA at a significantly lower extent than hAPE1. We consequently evaluated the endonuclease activity of zAPE1 on abasic dsDNA in comparison to that of hAPE1 and of the different K to A mutants in a set of dose-response experiments. As it is evident in Figure 7C, right, zAPE1 endonuclease activity was higher than that of hAPE1 and similar to that of the K to A and of the NΔ33 deletion mutants. Therefore, both the RNA-binding and the endonuclease activities of zAPE1 are different from that of hAPE1 and resemble those of the mutants we tested, strongly supporting the evolutionary-based hypothesis that the K31 and K32 evolutive insertion may represent a phylogenetic acquisition of a further point for functional regulation of the protein through PTMs. Domain swapping experiments are actually in progress to further substantiate this hypothesis.
Although some data suggest that hAPE1N-terminal region is important for recruiting proteins involved in BER processes (20) or in RNA metabolism (11), current opinion confines its value to the presence of the NLS sequence and of a pair of lysine residues (K6 and K7) whose acetylation controls the expression of some genes (8,32). Based on the experimental evidences that: (i) the truncated form of hAPE1 and its K to A mutants in the N-terminal domain, have higher endonuclease activity than the wild-type protein on abasic dsDNA (this study); (ii) hAPE1NΔ33 does not cleave 34FRNA (11); (iii) hAPE1NΔ33, as well as hAPE1 K5 mutant shows a significantly reduced ability to bind RNA molecules and NPM1 with respect to hAPE1 [(11) and this study]; (iv) the behaviour of K3 mutant on both binding to RNA and NPM1 and endonucleolytic activity is intermediate between that of the wild type and that of hAPE1NΔ33 (this study), we propose that the hAPE1 first 33 amino acids might have an important role in regulating its activities and the regulatory function of the K residues within the region 23–33 may require differential and non-mutually exclusive modulation.
hAPE1 full-length is endowed with distinct domains capable of binding oligonucleotides through electrostatic interactions (15). In addition to the positively charged residues within the catalytic domain that drive the binding of abasic nucleotides to the active site, the recognition of substrates and/or reaction product also involves residues within the 33N-term. Although positively charged residues are abundant in the 33N-term—eight lysines and two arginines conferring to the whole protein a net positive charge—results presented in this study clearly indicate that among them, a specific subset of lysine residues, located in the portion 23–33, plays an active role in hAPE1 catalysis.
The experimental data reported here on the role of the 33N-term in hAPE1 catalysis may be tentatively interpreted in structural terms by assuming that the binding of apurinic/apyrimidinic dsDNA is essentially driven by the residues of the catalytic site. On the other hand, the reduced affinity of the hAPE1 catalytic site for other DNA structures, such as reaction product or single stranded RNA/DNA molecules, makes the role of the 33N-term relevant. Its involvement in oligonucleotide recognition is not surprising, as Gly41 is not far (<8A) from a 15-bp DNA duplex containing an abasic site (Supplementary Figure S6), when this nucleotide is complexed with a truncated (NΔ40) form of APE1 (15). Within this framework, the decreased catalytic activity of hAPE1 on dsDNA compared to that exhibited by hAPE1NΔ33 should be ascribed to the contribution of the 33N-term to the electrostatic binding of the reaction product, being consistent with the increased activity shown by hAPE1K5 and with the effect of KCl concentration on the hAPE1 activity and not on that of the mutants hAPE1K5 and hAPE1NΔ33. The hypothesis is supported by other studies, which show that the catalytic activity of hAPE1 on abasic RNA/ssDNA (where the substrate binding is markedly dependent on the presence of the 33N-term) is strongly reduced >50mM KCl, whereas that on dsDNA is high even at 200mM (43–45).
Substrate recognition, cleavage and product release are separable molecular steps. The stimulation of DNA Pol β over hAPE1 activity, through elimination of the incision product, strongly suggests that release of the catalysis product is the limiting step in the removal of AP-sites by hAPE1 (15,46). Interestingly, Marenstein et al. (47) showed no product inhibition effect on the rate of hydrolysis of abasic ssDNA, while such an inhibition was observed for the corresponding dsDNA, thus suggesting that dsDNA product can make the difference in being more strongly bound to the N-terminal disordered arm. Altogether, our results point to the fact that acetylation of the critical lysine residues located in the 23–33 protein sequence and involved in inhibition of both NPM1 and RNA interaction, is also very important in modulating hAPE1 endonuclease activity, presumably by preventing the enzyme N-terminal region to assume an electrostatically stabilized conformation associated to a reduced endonuclease activity. Therefore, we can speculate that the N-terminal tail of hAPE1 may regulate the overall enzyme catalytic turnover, reducing the rate of product release without appreciably affecting the AP-site recognition specificity of the enzyme. The possibility that the product release step may be regulated in vivo by PTMs or interactions with other proteins, such as NPM1 (11), CSB (21) or XRCC1 (20), could be viewed in the light of the necessity to coordinate the BER activities on different substrates in order to protect the abasic DNA repair intermediate and to ensure its coordinated processing.
Notably, some of the lysine residues (i.e. K27, K31 and K32) involved in the interaction with NPM1 and with RNA and responsible for modulating the catalytic activity of the enzyme, were found acetylated in vivo (Figure 6), thus providing further biological relevance to our findings. Based on the evidence that mutations inserted to mimic constitutive acetylation (the K to A mutants) progressively increase the catalytic activity of hAPE1 on abasic dsDNA, the occurrence of such specific PTMs may well explain the observation made by different laboratories that immunopurified hAPE1 forms from mammalian cells retains significantly higher enzymatic activity than that of recombinant proteins expressed in E. coli. From our data we can also infer that simultaneous multiple acetylation is required to modulate the different hAPE1 known activities.
Taken together and in light of previous works (11,27,48), our current data allow to propose the molecular model of APE1 functions reported in Figure 8. After APE1 scans nucleic acids in search of its substrate and catalyzes the cleavage of the abasic site, acetylation of the critical lysine residues (i.e. K27, K31 and K32) should stimulate the overall hAPE1 endonuclease activity through increasing the product release rate. Specific acetylation may also modulate the interaction of hAPE1 with NPM1 (and possibly with other unidentified protein partners) and thus could redirect a functionally active form of hAPE1 from buffering storage sites, such as nucleoli, toward the active sites of DNA repair. Moreover, acetylation of the critical lysine residues may also inhibit hAPE1 action in RNA metabolism by preventing its RNA-binding activity. Notably, a recent paper provided the first evidence for ubiquitination occurring on K24, K25 and K27 (31). Interestingly, an ubiquitin-fused hAPE1 recombinant protein, in which the ubiquitin sequence is sandwiched between A23 and K24 (mimicking constitutively ubiquitinated hAPE1) was clearly excluded from nuclei (31). Thus, it can be speculated that post-translational modifications occurring on or near these critical lysine residues may play a significant role in controlling hAPE1 subcellular distribution, its functional activation and thus its biological function in mammalian cells.
From an evolutionary perspective, hAPE1 activity was made tunable (through post-translational modifications or the interaction with other proteins) or expanded toward other substrates with the acquisition of a positively charged N-terminus, without major modifications of its catalytic site, which retained the ‘canonical’ function towards abasic dsDNA. The existence of such an N-terminal extension for mammals only (Figure 7), may suggest its evolutionary significance in facing increased functional complexity. The intrinsic lack of structure of this domain can confer functional advantages to hAPE1, including the ability to bind to several different targets (i.e. NPM1, XRCC1, CSB, RNA, etc.), thus allowing precise control over the thermodynamics of the binding process (49). Since the 33N-term is required for both stabilizing the interaction to RNA and controlling the overall hAPE1 enzymatic activity through decreasing the rate of product release, this double role provides a ‘relay’ for molecular regulation with extremely useful biological significance and could also be envisioned in the light of preventing formation of harmful unprotected DNA strand breaks. Interestingly, through multiple interactions with different protein partners, the enzymatic activity of hAPE1 could be finely tuned, in association with the biological circumstances, in a time-dependent manner. While this work was in progress, a article by Yu et al. demonstrated that the K residues within the 24–35 region, which have a differential protection pattern between free and DNA-bound complexes of APE1, are involved in a conformational change concomitant with DNA binding and catalysis and with interaction with Pol β (50). This work and our current data strongly support the notion that these critical residues may indeed constitute a structural switch for hAPE1 functional regulation.
Regulation of the interaction of hAPE1 with NPM1 (11) may have profound biological consequences during cell response to stressor signals. NPM1 physically interacts with p53 and is believed to regulate this tumor suppressor protein (51). p53-signaling network, and its involvement of ubiquitin in particular, plays an important role in APE1 functional regulation, as recently demonstrated (31). During DNA damage stress response, ARF dissociates from NPM1 within nucleoli and relocalizes to the nucleoplasm where it sequesters the p53-ubiquitinating enzyme MDM2, thus leading to p53 up-regulation and functional activation (52). These data support the hypothesis that nucleolus may represent a major cellular stress sensor transmitting signals to the system for regulation of p53 activity (53). Thus, a careful elucidation of the p53-signalling networks and their involvement in the DNA repair pathway coordinated by hAPE1 is an important subject for molecular carcinogenesis and deserves further study. To this aim, work is now in progress in our laboratory to study the modulation of the hAPE1 interactome network during genotoxic damage and to evaluate its impact on the dynamics of nucleolar hAPE1 distribution. In conclusion, our work underscores the functional importance of the neglected unstructured hAPE1N-terminal domain and points to the need of structural details on this protein region for understanding the fine-tuning mechanisms regulating this multi-functional protein.
Supplementary Data are available at NAR Online.
MIUR (FIRB RBRN07BMCT_008, MERIT and PRIN 2008CCPKRP_002/003, to G.T. and A.S.); Galileo Project 2008 (to G.T. and J.P.R.); Association pour la Recherche sur le Cancer (ARC1118, to J.P.R.); EU-US bilateral project from Italian Ministry of Foreign Affairs (to G.T.). Funding for open access charge: MIUR (FIRB RBRN07BMCT_008 and PRIN 2008CCPKRP_002/003 (to G.T.).
Conflict of interest statement. None declared.
Authors thank Raphaël Guérois for his help with the interpretation of protein multiple sequence alignment, Dr Mark R. Kelley for zAPE1 expressing vector and Dr David Wilson III for critically reading the article and helpful suggestions.