Protein Phosphatase 2A-Dependent Dephosphorylation of Replication Protein A Is Required for the Repair of DNA Breaks Induced by Replication Stress

Abstract

Eukaryotic genomic integrity is safeguarded by cell cycle checkpoints and DNA repair pathways, collectively known as the DNA damage response, wherein replication protein A (RPA) is a key regulator playing multiple critical roles. The genotoxic insult-induced phosphorylation of the 32-kDa subunit of human RPA (RPA32), most notably the ATM/ATR-dependent phosphorylation at T21 and S33, acts to suppress DNA replication and recruit other checkpoint/repair proteins to the DNA lesions. It is not clear, however, how the DNA damage-responsive function of phosphorylated RPA is attenuated and how the replication-associated activity of the unphosphorylated form of RPA is restored when cells start to resume the normal cell cycle. We report here that in cells recovering from hydroxyurea (HU)-induced genotoxic stress, RPA32 is dephosphorylated by the serine/threonine protein phosphatase 2A (PP2A). Interference with PP2A catalytic activity causes persistent RPA32 phosphorylation and increased HU sensitivity. The PP2A catalytic subunit binds to RPA following DNA damage and can dephosphorylate RPA32 in vitro. Cells expressing a RPA32 persistent phosphorylation mimetic exhibit normal checkpoint activation and reenter the cell cycle normally after recovery but display a pronounced defect in the repair of DNA breaks. These data indicate that PP2A-mediated RPA32 dephosphorylation is required for the efficient DNA damage repair.

The genomes of all living cells are under constant attack by exogenous DNA-damaging agents, as well as the intracellular by-products of normal metabolism. To cope with this challenge, eukaryotic cells have evolved an elaborate surveillance and maintenance system termed the DNA damage response, which is composed of a set of signal transduction and execution pathways that can detect DNA lesions, delay cell cycle progression, facilitate repair processes, and induce apoptosis or senescence if the level of DNA damage is beyond repair (reviewed in references 18, 40, and 45).

The extensive phosphorylation of many checkpoint and DNA repair proteins by two phosphatidylinositol 3-kinase-related kinases (PIKKs) ATM and ATR, along with their respective preferred downstream kinases Chk2 and Chk1, appears to play a major theme in the transduction and execution of the DNA damage response. Once these kinases are stimulated, they phosphorylate and activate important regulators such as Rad17, Nbs1, BRCA1, H2AX, the 32-kDa subunit of replication protein A (RPA32), Cdc25, and p53 to facilitate assembly of DNA repair centers (foci) at the sites of DNA damage or cause alteration of their enzymatic or transcriptional activities leading to cell cycle arrest, apoptosis, or senescence (reviewed in references 1 and 26). Despite their overall similarity, these two pivotal pathways differ in the types of DNA damage to which they respond. While the ATM-Chk2 pathway responds primarily to DNA-damaging reagents that induce DNA double-stranded breaks (DSBs), the ATR-Chk1 pathway plays a predominant role in the cellular responses to UV and hydroxyurea (HU), which induce base damage or replication fork stalling (27, 43).

Although much is known about the role of protein phosphorylation in the DNA damage response, there is relatively little knowledge available concerning the function of protein dephosphorylation in this process. Recently, mounting evidence has prompted an emerging view that dephosphorylation of these phosphorylated checkpoint and repair proteins may also serve an important role in the DNA damage response, possibly by allowing cells to recover from checkpoint arrest or by facilitating the repair of DNA damage (reviewed in references 18 and 34). In Schizosaccharomyces pombe and Saccharomyces cerevisiae, recovery from checkpoint arrest following repair of DNA lesions may require dephosphorylation and inactivation of Chk1 or Rad53 (a yeast Chk2 orthologue), which is mediated by Dis2 (S. pombe), a member of the protein phosphatase 1 (PP1) phosphatase family, or Ptc2 and Ptc3 (the Wip1/PPM1D homologues) (S. cerevisiae), members of the PP2C family of phosphatases (14, 28). In humans, it appears that reentry into the cell cycle after the DNA damage response may depend heavily on Wip1/PPM1D, which can reportedly dephosphorylate Chk1, Chk2, and p53 (15, 31). In addition, BRCA1-dependent DNA DSB repair requires PP1-dependent dephosphorylation of BRCA1 (21, 57). Removal of γ-H2AX, the phosphorylated histone H2AX generated at the site of DNA DSBs whose function is to stabilize DNA repair foci, is mediated by PP2A or PP4 and functions to facilitate repair of DNA DSBs and allows subsequent resumption of DNA replication (9, 10).

Replication protein A (RPA) is a heterotrimeric protein complex composed of the 70-kDa, 32-kDa, and 14-kDa subunits (hereafter referred to RPA70, RPA32, and RPA14, respectively). RPA is a major single-stranded DNA (ssDNA)-binding protein in eukaryotic cells and is essential for all types of DNA metabolism, including DNA replication, DNA recombination, and DNA damage repair (53). Besides its protective role in covering the ssDNA exposed by DNA damage, RPA has also been shown to be critical in activation of the ATR pathway, likely by mediating the recruitment of ATR to the sites of DNA damage through the interaction between ATRIP and the RPA-ssDNA complexes (60). In addition, translocation of other important checkpoint/repair protein complexes, such as Rad17-Rfc2-5, Rad9-Hus1-Rad1, and Rad51/Rad52, to the sites of DNA lesions may also rely on RPA (16, 38, 54, 61).

RPA is regulated through genotoxic stress-induced phosphorylation. Upon DNA damage or replication stress, the SQ/TQ motif-containing threonine 21 (T21) and serine 33 (S33) of the RPA32 subunit are phosphorylated in a process mediated by PIKKs ATM, ATR, and DNA-dependent protein kinase (DNA-PK). In addition, phosphorylation of RPA32 is also seen on at least five other serine residues (S4, S8, S11, S12, and S13), but the identities of the responsible kinases remain to be determined (reviewed in references 6 and 62). It has been demonstrated that RPA32 phosphorylation plays an important role in the DNA damage response. Phosphorylation of RPA32 stimulates genotoxic stress-induced interaction with two critical checkpoint/repair complexes, Mre11-Rad50-Nbs1 and Rad9-Hus1-Rad1 and may also promote the recruitment of the DSB repair proteins Rad51/Rad52 to sites of DNA damage (39, 54, 55). While phosphorylated RPA is competent to translocate to locations of DNA lesions, it is unable to associate with replication centers and thus may function to mediate the S-phase checkpoint by suppressing DNA replication directly (36, 37, 48). The inability of phosphorylated RPA to support DNA replication may be due to its altered duplex DNA binding/denaturation ability and decreased interaction with DNA polymerase α (Pol α) (5, 35, 37). It is noteworthy that during this process, ATM/ATR-dependent phosphorylation of RPA at T21 and S33 is critical, whereas phosphorylation at other sites appears to be dispensable, indicating that distinct RPA functions may be differentially regulated by phosphorylation at different sites (36).

Given the essential roles that unphosphorylated RPA plays in unperturbed cell growth and division, it is conceivable that DNA damage-induced RPA32 phosphorylation needs to be attenuated when cells are recovering. Here we report that the ATM/ATR-dependent phosphorylation of RPA32 at T21 and S33 is reversed by PP2A-mediated dephosphorylation. Interference with PP2A activity causes persistent RPA32 phosphorylation and increased sensitivity to a replication stress inducer HU. The PP2A catalytic subunit associates with and dephosphorylates RPA32 following HU-induced stress. Through a model that mimics persistent phosphorylation of RPA32, we show that cells substituted with T21 and S33 (T21/S33)-phosphomimetic RPA32 exhibit increased HU/UV sensitivity but nonetheless possess normal checkpoint activation and indistinguishable resumption of DNA replication and progression through mitosis after HU release compared with wild-type (WT) control cells. Further investigation demonstrates that following release from HU treatment, cells with mutant RPA32 display persistent DNA damage foci containing RPA and γ-H2AX and exhibit a pronounced defect in the repair of HU-induced DNA breaks, suggesting that PP2A-mediated RPA32 dephosphorylation is required for the efficient repair of DNA lesions.

MATERIALS AND METHODS

Plasmids and mutagenesis.

C-terminally Flag-tagged RPA32 was generated by subcloning into the pcDNA3 vector a PCR product containing the RPA32 coding sequence from p3a-RPA32, a generous gift from Marc Wold (University of Iowa). The pcDNA3-RPA32-Flag was further used as template to introduce the T21V and S33A (T21V/S33A) and T21D and S33D (T21D/S33D) substitutions after several rounds of mutagenesis following the QuikChange site-directed mutagenesis protocol (Stratagene). The mutagenesis primers are shown in Table S1 in the supplemental material. The Flag-tagged RPA32 variants were subsequently cloned into pQCXIP (Clontech) vector for generating retroviruses. The retroviral RPA32 short hairpin RNA expression vector was described previously (36). The expression constructs of the Flag-tagged PP1, PP2A, PP4, and PP6 catalytic subunits were generously provided by Xin-Hua Feng (Baylor College of Medicine). The Flag-tagged PP5 expression construct was described previously (2).

Cell culture and retroviral infection.

HeLa and A549 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. U2OS cells were maintained in McCoy's 5A medium with 10% fetal bovine serum The culture of HaCaT and HepG2 cells has been described previously (49). Various Flag-tagged RPA32 variants were introduced into HeLa cells by retroviral infection, followed by selection of puromycin-resistant cells. The substitution lineages were further generated by silencing the endogenous RPA32 in these cell lines that ectopically express each of the mutant RPA32 proteins, which was accomplished by two rounds of infection with retroviruses expressing the RPA32 short hairpin RNA target sequence, followed by G418 selection. The retroviral infection protocol was described previously (17).

Antibodies and reagents.

The antibodies used for immunoblotting were purchased from Abcam (rabbit polyclonal RPA32pT21 and PP4), BD Biosciences (rabbit polyclonal PP5), Bethyl Laboratories (rabbit polyclonal RPA32pS33, RPA32pS4/8, and Rad17pS645), Calbiochem (mouse monoclonal RPA32 and RPA70), Cell Signaling (rabbit polyclonal γ-H2AX and Chk1 pSer317 and rabbit monoclonal Chk1 pSer345 and p21 Waf1/Cip1), Millipore Chemicon (rabbit polyclonal PP6), Upstate (rabbit polyclonal PP1 catalytic subunit and mouse monoclonal PP2A/Cα), Santa Cruz (mouse monoclonal p53), and Sigma-Aldrich (mouse monoclonal Flag and γ-tubulin). The reagents used in this study include HU, okadaic acid (OA), DNase, and caffeine (all from Sigma-Aldrich) and MG132 (BioMol).

Immunoblotting and immunoprecipitation.

For immunoblotting, cells were lysed in NETN (20 mM Tris-Cl [pH 7.5], 150 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40) supplemented with protease inhibitors (20 μg/ml leupeptin, 10 μg/ml pepstatin A, and 10 μg/ml aprotonin) and phosphatase inhibitors (20 mM β-glycerophosphate and 0.5 μM OA). Cell lysates were subsequently resolved by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and immunoblotted with appropriate antibodies. For examination of the interaction between PP2A/C and RPA32, HeLa cells were harvested with the previously described PP2A coimmunoprecipitation buffer (3) (except that it contained 0.5% Nonidet P-40) containing protease and phosphatase inhibitors. Lysates were precleared and incubated with the anti-RPA32 antibody and protein A beads (Calbiochem). After extensive washing with lysis buffer, the immunoprecipitates were analyzed by immunoblotting with anti-RPA70 and anti-PP2A/C antibodies. The control mouse immunoglobulin G was from Santa Cruz.

Immunofluorescence microscopy.

Cells plated on glass coverslips were treated with various conditions before immunofluorescence analysis. In brief, cells were fixed and permeabilized with acetone-methanol (1:1) at −20°C for 10 min. After rehydration in phosphate-buffered saline (PBS) and incubation with blocking buffer (3% bovine serum albumin [BSA] in PBS containing 0.1% Triton X-100) for 1 h, the cells were stained with primary antibodies in blocking buffer for 1 h at room temperature or overnight at 4°C, followed by three PBS washes and incubation with the appropriate secondary antibodies (Invitrogen) in blocking buffer for 2 h at room temperature. After the cells were washed with PBS, they were mounted with fluorescence mounting medium containing 4′,6′-diamidino-2-phenylindole (DAPI) (Santa Cruz) and analyzed by fluorescence microscopy. For detection of incorporated bromodeoxyuridine (BrdU), cells were incubated in medium containing 10 μM BrdU (BD Pharmingen) for 10 min prior to fixing and the DNA was denatured with 0.1 M HCl using standard procedures. The primary antibodies used were RPA32 and RPA70 (mouse monoclonal antibodies; Calbiochem), RPA32pS33 (rabbit polyclonal antibody; Bethyl), γ-H2AX (rabbit polyclonal antibodies [Cell Signaling] and mouse monoclonal antibody [Upstate]), PP2A/C (rabbit polyclonal antibody; Santa Cruz), and anti-BrdU (rat monoclonal antibody; Abcam). Experiments were performed three times, and cells detected with five or more discrete foci were regarded as focus positive.

Silencing of protein phosphatases.

To knock down OA-sensitive protein phosphatases (PP1, PP2A, PP4, PP5, and PP6), approximately 2 × 10⁵ HeLa cells were seeded per well in six-well plates. The next day, cells were transfected with small interfering RNA (siRNA) oligonucleotides against the catalytic subunit of the specific phosphatases (Santa Cruz) using Lipofectamine RNAiMAX reagent (Invitrogen) according to the manufacturer's instructions. Cells were treated with the indicated conditions and analyzed 48 to 72 h posttransfection.

In vitro phosphatase assay.

HeLa cells were treated with 10 mM of HU for 6 h, and the phosphorylated RPA32 substrate was prepared by immunoprecipitation using anti-RPA32 antibody and protein G beads. The phosphatase reactions were carried out in dephosphorylation buffer (20 mM HEPES [pH 7.0], 1 mM dithiothreitol, 1 mM MnCl₂, 100 mg/ml BSA, and 50 mM leupeptin) with purified PP2A enzyme (Upstate) at 30°C for 30 min in the presence of different concentrations of OA.

Clonogenic survival assays.

HeLa cells were plated at 500 cells per dish in 60-mm dishes and exposed to pulses of the indicated doses of HU or UV 12 h postseeding. After treatment, cells were rinsed twice with PBS and allowed to recover in drug-free medium. The cultures were then incubated for 10 to 14 days with the medium being changed every 3 days. Colonies were stained with crystal violet (Sigma), counted, and normalized to untreated control. All surviving points were done in triplicate, and only colonies containing 50 or more cells were scored.

DNA synthesis assays.

The DNA synthesis assays were performed as described previously (11) to evaluate the UV-induced intra-S-phase checkpoint and the resumption of DNA synthesis following release from HU block. Specifically, cells were incubated in medium containing 20 nCi/ml of [¹⁴C]thymidine (NEN) for 24 h prior to UV treatment or exposure to pulses of HU. At the indicated time points after UV irradiation or release from HU block, cells were pulse labeled with [³H]thymidine (2.5 μCi/ml, 30 min; NEN) and harvested. The radioactivity was determined by liquid scintillation counting, and the relative DNA synthesis rate was calculated by determining the ratio of ³H to ¹⁴C and normalization of the treated samples to the appropriate controls.

Cell cycle analysis.

Cells pulse exposed to HU were harvested at various recovery time points. After fixation in 70% ethanol, cells were incubated with RNase A (100 μg/ml; Invitrogen) and propidium iodide (50 μg/ml; Sigma) for 30 min at 37°C. The DNA content was then determined with a FACScan flow cytometer (BD Biosciences), and the cell cycle distributions were analyzed by CellQuest software.

Mitotic index assays.

The mitotic index assays were carried out to examine the UV-induced G₂/M checkpoint and the progression of mitosis following release from HU block. Specifically, cells were irradiated with UV or pulse treated with HU, and at the indicated time points following treatment, cells were harvested and fixed in 70% ethanol at −20°C. After permeabilization in 0.25% Triton X-100 in PBS, cells were incubated with anti-phospho-histone H3 antibody (pSer10; Upstate), followed by fluorescein isothiocyanate (FITC)-conjugated secondary antibody (Jackson ImmunoResearch). Cells were subsequently counterstained with propidium iodide, and the phospho-histone H3 fluorescence and DNA content were determined by flow cytometry. The percentage of mitotic cells was calculated as the mitotic index. For examination of the G₂/M checkpoint, the mitotic index was normalized to that in the unperturbed controls.

Sequential labeling with IdU and CldU.

The in vivo labeling of DNA with iododeoxyuridine (IdU) and chlorodeoxyuridine (CldU) was performed as previously described (58). Briefly, cells were grown on slides, fixed with cold methanol, stored at 4°C for more than 30 min, and incubated in 1.5 N HCl in 0.5% Triton X-100 for 30 min at room temperature. After the cells were blocked in 3% BSA in PBS, the slides were incubated with rat anti-BrdU antibody BU1/75 (it can detect CldU but does not bind IdU; AbCam) for 1 h at 37°C and then incubated for a second time with mouse anti-BrdU monoclonal antibody (BD Pharmingen) for 1 h at 37°C. After incubation with appropriate secondary antibodies, the slides were subjected to immunofluorescence.

Single-cell gel electrophoresis assay (comet assay).

The repair kinetics of HU-induced DNA breaks was evaluated by the alkaline comet assay according to the manufacturer's protocol (Trevigen). Briefly, cells were pulse exposed to HU (0.2 mM, 24 h) and harvested at various recovery time points for single-cell gel electrophoresis. Nuclei were stained with Sybr green, and comets were visualized by epifluorescence on a Zeiss microscope. Images were analyzed using the public domain software program ImageJ, and comets were evaluated by quantifying the tail moment of 75 cells using the comet-analyzing program CometScore (Tritek).

Senescence assay.

Cells were cultured in six-well plates and underwent appropriate treatments before they were evaluated for β-galactosidase activity according to the manufacturer's protocol (Calbiochem). The percentage of β-galactosidase-positive cells was determined by bright-field microscopy after 200 cells were scored for each sample.

Apoptosis assay.

The percentage of apoptotic cells in the lethally treated cells was analyzed by an apoptosis assay kit according to the manufacturer's manual (Invitrogen). Briefly, cells were collected, washed once with cold PBS, and resuspended in the annexin-binding buffer before they were incubated with FITC, annexin V, and propidium iodide for 15 min at room temperature. The stained cells were then analyzed by flow cytometry.

RESULTS

RPA32 undergoes dephosphorylation in cells released from HU block.

To investigate whether genotoxic insult-induced RPA32 phosphorylation is attenuated during the recovery process, we examined the kinetics of RPA32 phosphorylation in HeLa cervical carcinoma cells released from the genotoxic stress induced by HU block. A potent inhibitor of the ribonucleotide reductase, HU causes ribonucleotide depletion and leads to inhibition of DNA replication and subsequent accumulation of DSBs as a result of replication fork collapse. As shown in Fig. 1A, HeLa cells accumulated a high level of phosphorylated RPA32 immediately before release from HU block (0.2 mM, 24 h), as indicated by the mobility upshift of RPA32 protein as well as an increase in the intensity of bands detected by the phospho-specific antibodies against RPA32 (T21 and S33) in comparison with unperturbed cells. Following release from HU treatment, the level of phosphorylated RPA32 gradually reduced and dropped to near basal levels at 12 h postrecovery, suggesting that RPA32 might undergo dephosphorylation in this recovery process (Fig. 1A).

FIG. 1.

RPA32 undergoes dephosphorylation in cells recovering from HU block. (A) HeLa cells were treated with pulses of HU (0.2 mM) for 24 h, washed twice with PBS, and allowed to recover in drug-free medium. The levels of phosphorylated RPA32 were detected at the indicated recovery time points by immunoblotting. untr, untreated. (B and C) HeLa cells were released from HU block (0.2 mM, 24 h). The attenuation of RPA32 phosphorylation at T21/S33 was analyzed at 0 h or 9 h postrecovery in the absence (−) or presence (+) of MG132 (20 μM) (B) or different doses of OA (0 to 100 nM) or OA (50 nM) combined with caffeine (2 mM) (C). (D) A panel of five different cell lines (HeLa, HaCaT, HepG2, U2OS, and A549) were released from HU block (0.2 mM, 24 h) and then analyzed for RPA32 dephosphorylation at T21 at the indicated times after recovery.

It is possible that the observed decrease in the level of phosphorylated RPA was the result of ubiquitin-dependent proteasomal degradation of the phosphorylated RPA32. To investigate this possibility, the proteasome inhibitor MG132 was applied to HeLa cells exposed to pulses of HU. As shown in Fig. 1B, while the presence of MG132 effectively abrogated degradation of p53, a protein which is continuously targeted for proteasomal degradation in HeLa cells due to the presence of human papillomavirus E6 protein (41), the addition of MG132 did not block the decrease in the phosphorylated form of RPA32 or have any noticeable effect on total RPA32 protein levels, suggesting that phospho-RPA32 is not targeted for proteasomal degradation. We next tested whether the reduction of RPA32 phosphorylation at T21 and S33 could be inhibited by okadaic acid, a wide-spectrum inhibitor of serine/threonine protein phosphatases. As indicated in Fig. 1C, the presence of as low as 50 nM of OA showed a significant suppression effect, suggesting that dephosphorylation by an OA-sensitive protein phosphatase may account for the declining RPA32 phosphorylation in cells recovering from HU block. However, it is possible that OA exerted its effect not by inhibiting RPA32 phosphatases but by activating the kinases that phosphorylate RPA32 at T21/S33 (ATM/ATR/DNA-PK). We ruled out this possibility by demonstrating that OA still prevented disappearance of phosphorylated RPA32 at T21 and S33 in the presence of 2 mM of caffeine, a well-known inhibitor of ATM/ATR (Fig. 1C). It should be mentioned that although DNA-PK was previously thought to be less sensitive to caffeine, it was reported that 2 mM of caffeine could still efficiently inhibit DNA-PK-dependent phosphorylation of RPA32 at T21 (7).

To ensure that these observations from HeLa cells were not dependent on the cell line, we conducted similar recovery experiments on four other cell lines, and the same results were obtained as indicated by the decrease in T21 and S33 phosphorylation of RPA32 (Fig. 1D). To explore whether RPA32 dephosphorylation could occur in cells recovering from a genotoxic insult other than HU, HeLa and U2OS cells were both irradiated with UV light. Apparent RPA32 dephosphorylation at S33 was observed 24 h after exposure to UV light, indicating that this is a general phenomenon (see Fig. S1 in the supplemental material). Thus, in cells recovering from HU stress or UV exposure, RPA32 phosphorylation is attenuated through dephosphorylation by an OA-sensitive protein phosphatase.

PP2A mediates RPA32 dephosphorylation at T21 and S33 during recovery from HU stress.

Although widely held as a general inhibitor of PPP subfamily of serine/threonine protein phosphatases, OA has been demonstrated to inhibit some phosphatases, which include PP1, PP2A, PP4, PP5, and PP6, more potently than others (25). In order to identify which OA-sensitive protein phosphatase mediates RPA32 dephosphorylation, a RNA interference-based screening was performed by transfecting HeLa cells with siRNA oligonucleotides against the catalytic subunits of these five OA-sensitive protein phosphatases prior to HU pulse exposure. As shown in Fig. 2A, silencing of PP2A, but not other OA-sensitive phosphatases, significantly attenuated RPA32 dephosphorylation at T21 and S33 in cells recovering from HU stress, suggesting that PP2A is likely the phosphatase that targets RPA32. It is also noteworthy that although knockdown of PP5 dramatically downregulated RPA32 phosphorylation before release from HU (Fig. 2A), which was consistent with our previous report that PP5 plays a positive role in the ATR activation (59), dephosphorylation of RPA32 was nonetheless apparent (Fig. 2A), indicating that PP5 is unlikely to be involved in this process.

FIG. 2.

PP2A mediates RPA32 dephosphorylation at T21 and S33. (A) HeLa cells were transfected with a mock siRNA oligonucleotide or with oligonucleotides against the catalytic subunits of PP1, PP2A, PP4, PP5, and PP6. Twenty-four hours posttransfection, cells were pulse treated with HU (0.2 mM, 24 h), and RPA32 dephosphorylation at T21/S33 was compared at the indicated time points. (B) HeLa cells transfected with control siRNA or siRNA against PP2A/C were pulse treated with HU (0.2 mM, 24 h). At 0 to 12 h postrelease, the cells were fixed and costained with DAPI and anti-RPA32pS33 and anti-γ-H2AX antibodies. The DNA damage-inducible foci were visualized by immunofluorescence microscopy, and the percentage of phospho-RPA32 and γ-H2AX focus-positive cells was determined, normalized, and plotted. The error bars represent the standard deviations from three independent experiments. (C) Immunoprecipitation (IP) assays were performed in unperturbed cells or cells recovering from exposure to pulses of HU (0.2 mM, 24 h) at the indicated times after release using an anti-RPA32 antibody. The association of RPA32 with PP2A was detected by immunoblotting (IB) using an antibody against PP2A/C (α-PP2A/C). Binding of RPA70 was also examined as a loading control. The positions of the immunoglobulin light and heavy chains detected in the reaction are indicated by asterisks. untr, untreated; IgG, immunoglobulin G. (D) HeLa cells were mock treated (untreated [untr]) or exposed to HU (0.2 mM) for 24 h. DNA damage-inducible foci of RPA32 and PP2A were examined by immunofluorescence. (E) RPA32 was immunoprecipitated from HeLa cells exposed to HU (2 mM) for 24 h and used as substrate for the in vitro phosphatase assay in the absence (−) or presence (+) of purified PP2A catalytic subunit and different doses of OA (0 to 25 nM). Western blot analysis was performed 30 min later to determine the level of phosphorylated RPA32.

In agreement with this result, phosphorylated RPA32 foci persisted significantly longer in the PP2A/C-silenced cells, with 53.2% of cells being focus positive 12 h after release from HU block. This was in sharp contrast with the mock control cells that showed a significantly reduced number of foci and had only 18.5% of focus-positive cells at the 12-h time point (Fig. 2B). It is also noteworthy that phosphorylated RPA32 foci colocalized with γ-H2AX foci, which displayed a similar kinetics as its number decreased over time during unperturbed recovery and also exhibited longer persistency in the PP2A-silenced cells. These data are consistent with a previous report that PP2A mediates γ-H2AX dephosphorylation following pulse exposure to camptothecin, a topoisomerase I inhibitor (9).

It is possible that PP2A does not mediate RPA32 dephosphorylation directly but is somehow involved in this process indirectly. To assess this possibility, a coimmunoprecipitation assay was performed to examine whether PP2A could interact with RPA32 at endogenous levels. As shown in Fig. 2C, the PP2A catalytic subunit did not associate with RPA32 in unperturbed cells, but strong binding was observed when cells were exposed to HU stress. This binding decreased significantly after 9 h of recovery. In addition, DNase treatment did not abrogate the association between RPA32 and PP2A catalytic subunit, ruling out the possibility that their interaction is mediated by the chromatin bridge (see Fig. S2 in the supplemental material). Therefore, PP2A likely plays a direct role in the process of RPA32 dephosphorylation due to its induced association with RPA32 following genotoxic exposure. In support of this notion, immunofluorescence analysis also showed that the catalytic subunit of PP2A colocalized with RPA32 foci in an HU stress-inducible manner (Fig. 2D). To examine whether PP2A could dephosphorylate RPA32 directly, phosphorylated RPA32 immunoprecipitated from HU-stressed cells was incubated with recombinant PP2A catalytic subunit. PP2A readily dephosphorylated RPA32 in vitro, reducing levels of phosphorylated RPA32 within 30 min, and this process was markedly inhibited by the addition of 5 nM of OA (Fig. 2E), a concentration known to block PP2A activity under the assay condition. Taken together, these data indicate that PP2A mediates RPA32 dephosphorylation at T21 and S33 in cells recovering from HU-induced genotoxic stress.

RPA32 dephosphorylation at T21/S33 is dispensable for the checkpoint activation but is required for recovery from HU stress or UV irradiation.

As previously mentioned, the unphosphorylated form of RPA plays an essential role in supporting DNA replication and other DNA metabolism under normal cellular conditions. It is therefore plausible that when cells are recovering from genotoxic stress, PP2A-dependent RPA32 dephosphorylation may be required to reverse genotoxic stress-induced RPA32 phosphorylation, consequently restoring RPA32 to its normal condition. In support of this hypothesis, PP2A-silenced HeLa cells exhibit deficient RPA32 dephosphorylation at T21 and S33 after HU pulse exposure (Fig. 2A and B) and more importantly displayed significantly reduced cell viability following pulse exposure to HU stress (24 h) compared to cells treated with mock oligonucleotides (see Fig. S3A in the supplemental material).

Since PP2A is a versatile protein phosphatase targeting myriad substrates in numerous cellular processes, it is possible that the increased sensitivity to HU induced by PP2A knockdown is caused by PP2A-dependent processes other than attenuated RPA32 dephosphorylation. To better address whether RPA32 dephosphorylation is required for this process, we created a RPA32-T21D/S33D mutant (called DD hereafter) to mimic the persistent phosphorylation state of RPA32 at T21 and S33. The serine/threonine-to-aspartate conversion has been extensively utilized as phosphomimetic to study protein functions, and in many cases, the structures and activities of these phosphorylation-mimicking proteins are identical to those of the actual phosphoproteins (22, 48, 52). In order to avoid potential interference by the presence of endogenous WT RPA32, HeLa substitution cells were created in which the endogenous RPA32 was stably replaced with a nontargetable WT RPA32, the RPA32-DD mutant, or phospho-deficient RPA-T21V/S33A (VA) mutant. As shown in Fig. 3A, endogenous RPA32 expression was effectively silenced by retrovirus-mediated siRNA, and the levels of the exogenously expressed RPA32 variants were comparable to the level of endogenous RPA in the vector control cells. Cells expressing the DD mutant of RPA displayed normal morphology (data not shown) and a lightly slower growth rate than that of control cells (see Fig. S3B in the supplemental material). The ability of RPA32-DD to complex with other RPA subunits was also indistinguishable from other RPA variants (data not shown), consistent with a previous report on a RPA32 phosphomimetic which contained as many as eight substitutions (48). However, compared with this report (48), which showed incompetent association of the extensively substituted RPA32 mutants with the DNA replication centers, the RPA32-DD described here displayed no apparent deficiency in its colocalization with the DNA replication centers in the unperturbed cells (see Fig. S3C in the supplemental material), indicating that the RPA32-DD is sufficient to support DNA replication, a result that is not surprising given the relatively normal cytology exhibited by the DD cells. No other discernible phenotypes were found after 50 passages.

FIG. 3.

RPA32 dephosphorylation at T21/S33 is dispensable for the checkpoint activation but is required for recovery from HU stress. (A) The various cell lines with substitutions in RPA32 (S) were created by stably expressing in HeLa cells the empty vector (V) or Flag-tagged RPA32 variants (WT, VA, and DD) followed by retrovirally silencing endogenous RPA32 in these overexpression cells (O). RPA32 levels in these cell lines were examined by Western blot analysis. (B) WT and DD cells were treated with HU (5 mM). At 0 to 3 h postexposure, cells were fixed and stained with anti-RPA32 antibody. The RPA32 foci were visualized by immunofluorescence, and the percentage of focus-positive cells was determined, normalized, and plotted. Error bars represent the standard deviations from three independent experiments. (C) The RPA32 substitution cells were mock treated (−) or exposed (+) to HU (5 mM), and phosphorylation of Chk1 at S345 and Rad17 at S645 was examined by Western blotting 3 h later. (D) Various cells were irradiated with 25 J/m² UV, and 3 h later, the DNA synthesis rate was determined and normalized to that in untreated cells (Untr). (E) The RPA32 substitution cells were left untreated or irradiated with 25 J/m² UV. Later (1.5 h later), the cells were costained with propidium iodide and anti-phospho-H3 antibodies, and the mitotic fractions were determined by fluorescence-activated cell sorting analysis and normalized to that of unperturbed cells. (F) The RPA32 substitution cell lines were pulse exposed to 0 to 1.2 mM HU for 24 h and then allowed to recover in drug-free medium for 10 to 14 days. The cell viability of various lineages was analyzed by the clonogenic survival assay. All the data points in panels D, E, and F represent the means ± standard deviations (error bars) from three independent experiments, each performed in triplicate.

Since RPA translocation to sites of DNA damage has been demonstrated to be an early event in the DNA damage response and plays an important role in activating ATR-dependent checkpoint pathways, we examined whether cells with the RPA32-DD phosphomimetic mutant exhibited defective upstream checkpoint activation following genotoxic stress. After HU treatment, RPA32-DD formed punctuate foci in the nucleus that increased over time at a rate similar to that of the RPA32-WT, indicating that the RPA32-DD mutant translocates to chromatin normally upon DNA damage (Fig. 3B). In addition, ATR-dependent phosphorylation of two important checkpoint regulators, Chk1 and Rad17, displayed no distinguishable difference in the DD cells compared with other cells, suggesting that the RPA32-DD substitution does not affect ATR activation (Fig. 3C). The DD cells have an intact intra-S checkpoint as evidenced by a reduction in DNA synthesis following UV irradiation to an extent comparable to that of the WT and vector control lines, but in contrast with the phospho-deficient AA cells which maintained a relatively higher level of DNA synthesis (Fig. 3D). This observation confirms two previous reports that RPA32 phosphorylation at T21/S33 is required for the intra-S checkpoint (36, 48). In addition, the mitotic index of the DD cells, as measured by phospho-histone H3 staining, was reduced significantly upon UV exposure, to a level similar to that of other control cells (Fig. 3E), indicative of a normal G₂/M checkpoint in the DD cells. These data suggest that substitution of endogenous RPA32 with the RPA32-T21D/S33D phosphomimetic mutant does not affect the checkpoint activation in the DNA damage response. Therefore, for studies of the recovery process from the DNA damage response where RPA32 is dephosphorylated at T21 and S33, the cells with the RPA32-DD substitution provide us with an opportunity to investigate the events influenced by RPA32 dephosphorylation by serving as a persistent phosphorylation mimic model.

To examine whether RPA32 dephosphorylation is required for efficient recovery from HU stress, the cells with substitutions in RPA32 were pulse treated to a range of doses of HU for 24 h, and their viability after recovery was analyzed by the clonogenic survival assay. As shown in Fig. 3F, the survival rate of DD cells was significantly reduced compared to that of WT or vector control cells; only 30.9% of DD cells survived and formed colonies after recovery from pulse exposure to 0.2 mM of HU, which was in sharp contrast with WT cells (72.4%) and vector control cells (70.8%). Similarly, DD cells were also less capable to survive UV irradiation compared to control cells, albeit to a lesser extent than with HU treatment (see Fig. S3D in the supplemental material). Taken together, cells with the RPA32-DD mutant are defective in the recovery from HU stress or UV irradiation, suggesting that RPA32 dephosphorylation at T21 and S33 is required in this process.

RPA32 dephosphorylation is not required for the resumption of DNA replication or subsequent initiation of mitosis following release from HU block.

It has been reported that genotoxic stress-induced RPA32 phosphorylation prevents its association with replication centers and thus may suppress DNA synthesis following DNA damage (48). Recent evidence indicates that phosphorylation at the ATM/ATR-responsive T21 and S33 sites is critical for this function (36). Given these findings, it is plausible that RPA32 dephosphorylation at T21/S33 might be required for the resumption of DNA replication following recovery from HU block. However, as the thymidine incorporation assay indicated, the DNA synthesis activity of the RPA32-DD substitution cells increased at a rate comparable to that of WT and vector control lines within 3 hours after release from HU pulse exposure (0.2 mM, 24 h) (Fig. 4A), implying that RPA32 dephosphorylation is not required to resume the arrested DNA replication following release from HU block.

FIG. 4.

RPA32 dephosphorylation is not required for the resumption of DNA replication or subsequent mitosis following release from HU block. (A) The RPA32 substitution cells were released from HU exposure (pulses) (0.2 mM, 24 h). At 0 to 3 h postrelease, the DNA synthesis rates were determined and normalized to the rate at 0-h time point. All samples were tested in triplicate, and consistent results were obtained among three independent experiments. (B) Various cell lineages were allowed to recover from pulse exposure to HU (0.2 mM, 24 h). Cells at the indicated time points were fixed and costained with propidium iodide (PI) and anti-phospho-H3 antibody, and the mitotic fractions were further determined by fluorescence-activated cell sorting (FACS) analysis. (C and D) HeLa cells were treated with sublethal (0.2 mM) or lethal (2 mM) doses of HU for 24 h and then allowed to recover. At 0 to 24 h after release, cells were harvested and analyzed by immunoblotting using appropriate antibodies (C). The remaining cells underwent cell cycle profiling analysis by PI staining and FACS (D). (E) HeLa cells were pulse exposed to lethal doses of HU (2 mM, 24 h). At 6 h postrecovery, cells were pulse-labeled with BrdU for 10 min followed by immunofluorescence analysis using anti-BrdU and anti-RPA32pS33 antibodies. The cell positive for both types of staining were boxed and magnified for clearer visualization. (F) HeLa cells were allowed to recover from HU treatment (0.2 or 2 mM, 24 h) (pulses), and mitotic index analysis was performed by FACS at the indicated time points after release.

HU is a reversible inhibitor of DNA replication, and cells synchronized by HU exposure reenter the cell cycle if released from HU block. Since the resumption of DNA replication seems to be unaffected, it is possible that progression through the downstream G₂/M boundary requires RPA32 dephosphorylation at T21/S33. To test this possibility, the various substitution lines were released from pulse exposure to HU (0.2 mM, 24 h), and their progression through mitosis was evaluated by phospho-histone-H3 staining at different time points. As indicated in Fig. 4B, the mitotic profile of the DD cells showed no noticeable difference compared with that of other two lines (VA and WT); the mitotic indices of all three lines remained low until 4 h postrelease when cells started to divide, and the percentage of the mitotic cells peaked at 8 h and dropped to a level similar to that of the 0-h time point at 10 h. Thus, RPA32 dephosphorylation at T21/S33 is not necessary for mitotic progression in cells recovering from HU block.

The above results were supported by an independent experiment performed on HeLa cells pulse exposed to two different levels of HU for 24 h. In sharp contrast with cells pulse treated with a sublethal dose (0.2 mM) of HU, where RPA32 was apparently dephosphorylated gradually during recovery, cells treated with a lethal dose (2 mM) exhibited persistent RPA32 phosphorylation up to 24 h postrelease (Fig. 4C). The lethally treated cells could resume DNA replication in S phase and reenter the cell cycle with no apparent defects, except that they displayed an approximately 3-h delay in their recovery compared with sublethally treated cells (Fig. 4D), indicating a dispensable role of RPA32 dephosphorylation in the resumption of DNA replication and the subsequent mitosis following release from HU block.

There is a possibility that DNA replication resumes only in cells where RPA32 is dephosphorylated, or alternatively at sites where the bound RPA32 is dephosphorylated and not at sites where RPA32 remains phosphorylated. To test these possibilities, HeLa cells were pulse-labeled with BrdU for 10 min 6 hours after release from HU block (2 mM, 24 h) and then tested for colocalization between phosphorylated RPA32 and incorporated BrdU, which usually indicates where DNA replication is occurring. As shown in Fig. 4E, phosphorylated RPA32 and BrdU strongly colocalized, indicating again that resumption of DNA replication is independent of RPA32 dephosphorylation. However, it is possible that DNA replication following release from HU block may not be resumed at the stalled replication forks but restart at the later origins that have not yet been fired, which could in theory allow for completion of DNA replication. To differentiate these two events, a previously reported sequential labeling approach (24, 58) was adopted such that cells were pulse-labeled with IdU before HU stress, followed by a second labeling with CldU at different time points after release from HU block. As indicated (see Fig. S4 in the supplemental material), all cells displayed a significant degree (>95%) of coincident labeling (yellow) of CldU (red) with prelabeled IdU (green) at 3 h postrelease from 0.2 mM of HU pulse treatment (HeLa and DD cells) or 6 h postrelease from 2 mM of HU pulse treatment (HeLa cells). In the latter case, RPA32 still remains hyperphosphorylated (Fig. 4C). In sharp contrast, only a negligible level (<5%) of postlabeled CldU (red alone) was observed. Thus, cells preferentially resume DNA replication at the previously stalled replication forks regardless of the phosphorylation status or the phosphomimetic mutation of RPA32, corroborating that RPA32 dephosphorylation is dispensable in resumption of DNA replication.

To confirm that RPA32 dephosphorylation is not required for resumption of mitosis following recovery from HU block, mitotic index analysis was carried out. As revealed in Fig. 4F, the lethally treated HeLa cells exhibit a lagging but nonetheless apparent progression through mitosis in comparison with the cells pulse exposed to a sublethal dose of HU, indicating that RPA32 dephosphorylation at T21/S33 is not required for this process. Taken together, these data indicate that RPA32 dephosphorylation is not required to resume DNA replication or progress through the mitotic phase during recovery from HU-induced genotoxic stress.

It is noteworthy that the kinetics of γ-H2AX dephosphorylation were similar to the kinetics of RPA32 dephosphorylation in cells pulse treated with sublethal doses of HU, and similar to RPA32, this dephosphorylation was significantly attenuated in lethally treated cells (Fig. 4C). In contrast, dephosphorylation of Chk1 at S345, a site whose phosphorylation has been shown to be critical for Chk1 activation (8), occurred under both conditions, albeit with a 3-h delay in lethally treated cells (Fig. 4C). It is also noteworthy that Chk1 dephosphorylation showed a strong temporal correlation with the release from HU block in both cases (Fig. 4C and D). These results are not surprising, since γ-H2AX colocalizes with phosphorylated RPA32 (Fig. 2B), which are both dephosphorylated by PP2A (9), whereas Chk1 dephosphorylation is reportedly mediated by a different protein phosphatase Wip1/PPM1D (31). Given the pivotal role Chk1 plays in checkpoint activation, it seems that Chk1 deactivation through dephosphorylation may also dictate checkpoint recovery from genotoxic stress. This idea is supported by a recent report that showed reversal of intra-S and G₂/M checkpoint activation to require PPM1D-dependent dephosphorylation of Chk1 (31).

RPA32 dephosphorylation at T21/S33 is critical for repair of DNA breaks induced by replication stress.

Because RPA has been shown to be involved in various DNA repair pathways, it is possible that the deficient recovery from HU stress and UV irradiation displayed by cells containing the RPA32 phosphomimetic mutant might be due to defective DNA damage repair. To investigate this possibility, the cells with substitutions in RPA32 were pulse exposed to HU (0.2 mM, 24 h) and then analyzed for the level of DNA breaks by alkaline single-cell gel electrophoresis (alkaline comet assay) at different recovery time points. Even though DD cells accumulated slightly lower levels of DNA breaks before HU release (data not shown), repair of the DNA breaks, as indicated by the decline of the relative comet tail moments, was significantly less efficient compared to WT cells (Fig. 5A). At 12 h postrelease, the DD cells still contained a substantially higher level of unresolved DNA breaks (22.4%) compared to the WT cells (3.3%). These data indicate that cells expressing the RPA32 phosphomimetic mutant are less efficient in the repair of DNA breaks induced by HU, suggesting that RPA32 dephosphorylation at T21/S33 is required for this repair process. Consistent with this result, PP2A/C-silenced HeLa cells exhibit less RPA32 dephosphorylation and a significantly higher level of unresolved DNA breaks 12 h after recovery from HU stress (36.7% relative to the 0-h point, calculated by the ratio between comet tail moments, Fig. 5B and C). This is in sharp contrast to the mock control cells where RPA32 was fully dephosphorylated and DNA repair was essentially complete (1.4%) (Fig. 5B and C).

FIG. 5.

RPA32 dephosphorylation is necessary for the efficient repair of DNA breaks induced by HU. (A) WT and DD cells were allowed to recover from HU exposure (0.2 mM, 24 h) (pulses). At 0 to 12 h postrecovery, cells were harvested and analyzed by alkaline comet assay. Repair of DNA breaks was evaluated by the level of the residual DNA breaks, which is calculated by the comet tail moments at various recovery time points relative to that at the 0-h time point. (B and C) HeLa cells were transfected with control siRNA or siRNA against the PP2A catalytic subunit. At 24 h posttransfection, the cells were pulse exposed to HU (0.2 mM, 24 h) and then allowed to recover. RPA32 dephosphorylation was analyzed by immunoblotting (B), and repair of DNA breaks was evaluated by alkaline comet assay (C). untr, untreated; RNAi, RNA interference. (D) WT and DD cells were allowed to recover from HU treatment (0.2 mM, 24 h) (pulses). At 0 to 12 h postrelease, cells were fixed and stained with DAPI and anti-RPA70 and anti-γ-H2AX antibodies. Immunofluorescence was carried out to visualize the DNA damage foci formed by RPA70 and γ-H2AX, and the percentages of the focus-positive cells were calculated, normalized, and plotted. All the data points in panels A, C, and D represent the means ± standard deviations (error bars) from three independent experiments.

In response to genotoxic stress, a large number of checkpoint and repair proteins including the Mre11/Rad50/Nbs1 complex, RPA, 53BP1, and γ-H2AX accumulate at sites of DNA damage and form huge protein complexes termed DNA damage foci, which are essentially the DNA repair centers that do not resolve until the repair process is completed (13, 30). For this reason, these protein foci are regarded as markers for DNA damage. To confirm the above results through the comet assays, we examined the kinetics of RPA foci and γ-H2AX foci, two markers broadly used to indicate DNA damage, in the cells with substitutions in RPA32 recovering from HU block (0.2 mM, 24 h). As shown in Fig. 5D, both RPA70 and γ-H2AX exhibited HU stress-induced punctate staining in WT and DD cells, which also displayed a significant level of colocalization. In both RPA32 substitution cells, the number of the speckles formed by these two proteins as well as the percentage of the focus-positive cells declined over time following recovery from HU pulse exposure. However, the rate of reduction in DD cells was significantly lower than that in WT cells; at 12 h postrecovery, only 18.7% of WT cells but as many as 46.6% of DD cells were γ-H2AX focus positive, and the percentage of DD cells that contained unresolved RPA70 foci was almost three times that of WT cells (29.2% versus 10.2%), confirming again that DNA damage repair is defective in the DD cells.

Taken together, these results indicate that the phosphomimetic DD mutant cells are deficient in the repair of DNA breaks induced by replication stress, demonstrating that PP2A-dependent RPA32 dephosphorylation at T21 and S33 is necessary for efficient DNA damage repair.

DISCUSSION

In this report we show that in cells recovering from the genotoxic stress induced by HU, RPA32 undergoes PP2A-dependent dephosphorylation at T21 and S33, a process dispensable for checkpoint activation or cell cycle reentry but critical for efficient DNA damage repair. Our findings, combined with several previous studies, have established that dephosphorylation of checkpoint/repair proteins is likely to be as important in the DNA damage response as the well-documented upstream phosphorylation cascades known to initiate and propagate the DNA damage response.

RPA is a versatile protein playing numerous roles in the DNA damage response, including recognizing and stabilizing DNA breaks, activating/mediating checkpoint pathways, and assisting with the downstream repair of diverse types of DNA lesions. Although much is known about the functions of RPA, little is known about how RPA's multiform activities are regulated. On the basis of our study and others, it seems that the dynamic phosphorylation and dephosphorylation of the RPA32 subunit at T21 and S33 might provide a mechanism controlling the numerous functions of RPA in the DNA damage response. Upon DNA damage or replication stress, RPA32 is phosphorylated by ATM/ATR at sites of DNA damage or stalled replication forks, and phosphorylated RPA suppresses ongoing DNA replication and helps recruit other checkpoint/repair proteins. Later, during recovery from DNA damage-induced checkpoint arrest, phosphorylated RPA32 is removed through PP2A-dependent dephosphorylation, an event that is essential for efficient DNA repair. For this entire process, initial phosphorylation and subsequent dephosphorylation of RPA32 need to be tightly regulated in order for a proper DNA damage response to occur. Suppression or failure of initial RPA32 phosphorylation causes a defective S-phase checkpoint leading to DNA damage-resistant DNA synthesis, and delay or dysfunction of subsequent RPA32 dephosphorylation results in inefficient repair of DNA lesions.

It is noteworthy that in addition to the ATM/ATR-responsive T21 and S33, several other serine residues (S4, S8, S11, S12, and S13) on the N terminus of RPA32 also undergo DNA damage-inducible phosphorylation. Therefore, it is possible that dephosphorylation of these sites is also regulated by PP2A and required for the DNA damage response. On the basis of our observations, PP2A is seemingly responsible for dephosphorylation of some, if not all, of these residues, since Western blotting with an anti-RPA32 antibody showed an apparent PP2A-dependent reduction of multiple slower-migrating bands in cells recovering from HU stress, which can be indicative of a loss of protein phosphorylation (Fig. 1A and 5B). In support of this, the results of Western blotting with an antibody that recognizes doubly phosphorylated RPA32 on both S4 and S8 suggested at least one of these two sites may also be dephosphorylated in a PP2A-dependent manner (see Fig. S5 in the supplemental material). Due to lack of appropriate phospho-specific antibodies against all of these seven individual sites, we did not address this question further. Regardless of whether PP2A is involved, given the high level of cytotoxicity induced by overexpression of RPA32 hyperphosphomimic mutant that has substitutions of all seven serine/threonine sites (data not shown) and the essential role unphosphorylated RPA32 plays in supporting unperturbed DNA replication, it is conceivable that cells would not survive persistent RPA32 hyperphosphorylation following genotoxic stress and that dephosphorylation of these DNA damage-inducible sites is likely a required event during recovery.

With the use of RPA32 phosphomimetic mutants, it has been shown by two groups that PIKK-dependent RPA32 phosphorylation may prevent RPA from associating with replication machinery following genotoxic stress, because these ectopically expressed RPA32 mutants exhibit defects in binding with replication centers and forming S-phase-specific foci (36, 48). In our study, however, the cytology of HeLa cells with the T21D/S33D mutant substitution was rather normal, and the DD mutant form of RPA32 displayed no apparent dysfunction in its colocalization with BrdU-labeled replication centers (see Fig. S3C in the supplemental material). One possible explanation for this is that RPA32-DD could still associate with replication centers but with slightly lower efficiency compared to wild-type RPA32. This could result in insufficient binding of RPA32-DD to replication centers if there is abundant endogenous wild-type RPA32 to compete with, as was the case in the two previous reports which both used overexpression systems (27, 29). If the interference caused by endogenous RPA32 is removed by RNA interference silencing, as in our substitution model, mutant RPA32 is still able to support DNA replication with no apparent defects. In support of this hypothesis, the DD substitution cells exhibit relatively normal replication, and the RPA32-DD mutant protein migrates to sites of DNA breaks efficiently and supports ATR-dependent checkpoint activation with no observable defects. Therefore, it is a useful model to study events downstream of checkpoint activation in the DNA damage response that are related to RPA32 dephosphorylation.

Surprisingly, our data indicate that RPA32 dephosphorylation is not required for resumption of DNA replication after release from HU block, as cells released from a lethal dose of HU resumed DNA replication normally (Fig. 4D and E) (see Fig. S4 in the supplemental material) regardless of persistent RPA32 phosphorylation (Fig. 4C). This result differs from the consensus view that phosphorylated RPA is unable to support DNA replication due to its reduced capability to bind with DNA Pol α and to denature duplex DNA (5, 35, 37). One possibility is that DNA synthesis during unperturbed S-phase replication and in the recovery from replication stress may utilize different DNA polymerases and involve distinct mechanisms. While the former process requires Pol α primase and unphosphorylated RPA-dependent initiation at replication origins and relies heavily on two other processive, high-fidelity polymerases (Pol δ and Pol ɛ) at replication elongation, DNA synthesis in the recovering cells mainly resumes at the arrested replication forks or starts as part of a repair process and therefore may not need the initiation step and may be mediated by nonreplicative but repair-specific polymerases (23, 42, 51). In support of this idea, several nonreplicative DNA polymerases, including X-family polymerases λ, μ, TdT, Y-family polymerase η, B-family polymerase ζ, and Rev1, have been found to be involved in the nonhomologous end joining and homologous recombination repair synthesis of DSBs, the major type of DNA breaks induced by collapse of replication forks following HU exposure (12, 19, 33, 42, 51). In addition, it has also been shown that the Pol α-primase complex is not required for DSB-induced homologous recombination during the mating type switch in budding yeast (50).

As discussed above, although resumption of DNA replication does not require RPA32 dephosphorylation, there was a close temporal correlation between resumption of DNA replication and Chk1 dephosphorylation regardless of the level of HU exposure (Fig. 4C and D). On the basis of a previous report that PPM1D-dependent dephosphorylation of Chk1 abrogates checkpoints, we believe that Chk1 dephosphorylation-dependent deactivation may dictate resumption of DNA replication as part of a checkpoint recovery process (31). Interestingly, while cells pulse treated with a lethal dose (2 mM) of HU and cells with the phosphomimetic mutation of RPA32 (DD) as well underwent apparent checkpoint recovery by resuming cell cycle following release (Fig. 4D) (see Fig. S7 in the supplemental material), these cells contained significantly high levels of unresolved DNA breaks during the recovery process, as indicated by the persistently high levels of phosphorylated RPA32 and γ-H2AX (Fig. 4C and 5D), as well as the data from the comet assays (Fig. 5A) (see Fig. S6 in the supplemental material). This is similar to an intriguing mechanism termed “checkpoint adaptation” originally discovered in yeast. In this mechanism, cells reenter the cell cycle despite the persistence of unrepaired DNA breaks, an advantage presumably conferred on unicellular yeast cells to allow them to survive a long-term arrest when facing persistent genotoxic stresses (47). This mechanism had been believed to be absent in metazoans because of the possibility that it could promote genomic instability, but recent studies have shown that such a phenomenon may exist in cells of higher organisms as well (18, 46, 56). These data in our study evidently suggest that checkpoint adaptation also occurs in human cells. Thus far, it is unclear why adaptation exists in higher organisms. Although it was speculated that this mechanism may facilitate the elimination of cells containing irreparable DNA damage through mitotic catastrophe (4), we observed a significant proportion of these adapted cells undergoing senescence, which was accompanied by a increase of p21 protein levels (see Fig. S8A and S8B in the supplemental material). Therefore, checkpoint adaptation may also ultimately trigger senescence. Although we did not see a significant contribution of apoptosis in this process (see Fig. S8C in the supplemental material), it is in theory possible that checkpoint-adapted cells may also end up being apoptotic. More in-depth studies are required to examine this interesting phenomenon.

This study has established a causal relationship between RPA32 dephosphorylation and repair of DNA breaks; however, the question of what specific role RPA32 dephosphorylation plays in this process remains open. Presumably, RPA32 dephosphorylation may play a role by recruiting certain repair-essential proteins to or removing other suppressive ones from the sites of DNA breaks where the DNA repair process is occurring. Accumulating evidence has shown that the repair of DNA breaks is a complicated multistep process orchestrated by a variety of DNA repair proteins that form dynamic but tightly regulated repair foci at sites of DNA damage (20, 29, 32, 44). Alternatively, phosphorylated RPA32 may negatively regulate certain downstream steps of DNA repair pathways that do not involve translocation of any repair proteins, and activation of these steps may therefore necessitate prior PP2A-dependent dephosphorylation of RPA32. To understand what specific role RPA32 dephosphorylation plays in the DNA repair process, it would be inviting to temporally dissect the entire process, including recruitment/removal of the repair proteins and activation of the stepwise DNA repair pathways.