Role of c-Abl Kinase in DNA Mismatch Repair-dependent G2 Cell Cycle Checkpoint Arrest Responses*

Current published data suggest that DNA mismatch repair (MMR) triggers prolonged G2 cell cycle checkpoint arrest after alkylation damage from N-methyl-N′-nitro-N-nitrosoguanidine (MNNG) by activating ATR (ataxia telangiectasia-Rad3-related kinase). However, analyses of isogenic MMR-proficient and MMR-deficient human RKO colon cancer cells revealed that although ATR/Chk1 signaling controlled G2 arrest in MMR-deficient cells, ATR/Chk1 activation was not involved in MMR-dependent G2 arrest. Instead, we discovered that disrupting c-Abl activity using STI571 (Gleevec™, a c-Abl inhibitor) or stable c-Abl knockdown abolished MMR-dependent p73α stabilization, induction of GADD45α protein expression, and G2 arrest. In addition, inhibition of c-Abl also increased the survival of MNNG-exposed MMR-proficient cells to a level comparable with MMR-deficient cells. Furthermore, knocking down GADD45α (but not p73α) protein levels affected MMR-dependent G2 arrest responses. Thus, MMR-dependent G2 arrest responses triggered by MNNG are dependent on a human MLH1/c-Abl/GADD45α signaling pathway and activity. Furthermore, our data suggest that caution should be taken with therapies targeting c-Abl kinase because increased survival of mutator phenotypes may be an unwanted consequence.

DNA mismatch repair (MMR) 3 proteins detect and repair mismatched bases or unpaired loops in DNA. Defects in MMR, primarily due to acquisition of mutations in both copies of the human (h) MLH1 (mutL homolog-1) or MSH2 (mutS homolog-2) genes (1)(2)(3), are directly linked to hereditary nonpolyposis colon cancer. A subset of sporadic colorectal cancers also lack MMR due to loss of hMLH1 protein expression caused by promoter hypermethylation (4). Cells deficient in MMR present work defects in G 2 cell cycle checkpoint arrest responses and display increased resistance to the lethal effects of specific DNA-damaging agents, such as N-methyl-NЈ-nitro-N-nitrosoguanidine (MNNG), 6-thioguanine (6-TG), 5-fluoro-2Јdeoxyuridine (FdUrd), cisplatin, and temozolomide (5). Theoretically, MMR-dependent G 2 arrest responses allow time for cells to repair mutagenic lesions created by these agents prior to cell entry and transit through mitosis. On several levels, MMR functions as a potent mutational avoidance system. Conversely, cells lacking MMR have a mutator phenotype as a result of failure to detect DNA lesions and subsequent absence of G 2 arrest and apoptotic responses (6,7), resulting in a greatly increased colon cancer risk (8).
MNNG causes a spectrum of specific DNA lesions, including methylation of dG at the O 6 -position in dG:dC base pairs. These DNA lesions are mutagenic because replication over O 6 -methylguanine base pairs causes preferential pairing with Thy. Fortunately, mutagenic O 6 -methylguanine:Thy mispairs are excellent MMR substrates (9); however, the exact signal transduction processes that regulate G 2 arrest responses remain ill defined. It is not clear if MMR signals arrest and cell death directly or if DNA replication is required in the context of newly synthesized DNA for signaling and cell death via DNA double-strand breaks (DSBs).
Two main theories have been proposed to explain MMR-dependent G 2 arrest and lethality responses to specific DNA damage in human cancer cells (6). In the futile cycle theory, MMR detects DNA lesions and creates DSBs due to futile cycles of repair or repair-replication fork collision. However, blocking signaling events emitted from MMR-lesion complexes may affect cell cycle and apoptotic "cleanup processes," but not lethality resulting from unrepaired DSBs. In the signaling theory, MMR acts as a DNA damage sensor, directly signaling G 2 arrest and cell death. Preventing these signaling pathways would ablate G 2 arrest and apoptosis and cause damage tolerance.
Redundant pathways promote G 2 arrest after DNA damage in human cells. G 2 arrest can be mediated by DNA damage activation of ATM (Ataxia telangiectasia mutated) or ATR (Ataxia telangiectasia-Rad3-related kinase) phosphatidylinositol 3-like kinases. These kinases phosphorylate and activate p53 and downstream Chk1 and Chk2 checkpoint kinases. Activated Chk1 or Chk2 can phosphorylate the Cdc25C phosphatase at Ser 216 , thereby inhibiting Cdc25C activity by 14-3-3-mediated cytoplasmic sequestration. This prevents Cdc25C from dephosphorylating Cdc2 at Tyr 15 , a main factor preventing the initiation of mitosis in mammalian cells. The phosphorylationmediated functional activation of the p53 tumor suppressor by the ATM/Chk2 and/or ATR/Chk1 pathway leads to its transcriptional regulation of several downstream genes involved in cell cycle checkpoint arrest, including p21, GADD45␣ (growth arrest-and DNA damage-inducible-45␣), and 14-3-3 (10). MMR-dependent GADD45␣ protein increases, in particular, were shown after FdUrd exposure (5), and GADD45␣ can mediate G 2 arrest by direct binding to and inhibiting Cdc2 (10).
Despite numerous indirect studies suggesting activation of specific pathways by MMR that might control G 2 arrest, a detailed analysis of MMR-dependent versus MMR-independent G 2 arrest responses has not been completed. MMR was reported to preferentially stimulate the ATR/Chk1 or ATM/ Chk2 pathway in cells after exposure to methylating agents such as 6-TG, MNNG, and temozolomide, the clinical equivalent to MNNG (11). p38␣ MAPK was also linked to MMRmediated G 2 arrest after temozolomide treatment. Inhibition of the proposed hMLH1/p38␣ pathway, by small interfering RNA (siRNA) or p38␣ inhibitors, abrogates MMR-dependent G 2 arrest (12). However, activation of this pathway appears to be independent of either Chk1 or Chk2 phosphorylation/signaling. Exposure of human cells to cisplatin can also stimulate MMR-dependent activation of the c-Abl/p73␣ pathway, leading to apoptosis (13). However, a link between MMR-dependent c-Abl kinase activation and G 2 arrest was not examined.
Using MMR-deficient human RKO colon cancer cells (lacking hMLH1 protein expression) corrected by single replacement and expression of hMLH1 protein at endogenous wildtype (WT) levels, we found that the ATR/Chk1 pathway was activated after MNNG treatment; however, its downstream phosphorylation of Chk1 was not MMR-dependent because equivalent consensus site Chk1 phosphorylation was noted regardless of MMR status. Furthermore, expression of dominant-negative ATR in WT MMR cells did not affect G 2 arrest or cell death responses, indicating that MMR does not signal through ATR.
Instead, we provide data in this and our accompanying article (41) showing that MMR-dependent G 2 arrest is controlled by hMLH1/c-Abl/GADD45␣ signaling. Although MNNG stimulated both ATR/Chk1 and hMLH1/c-Abl/GADD45␣ signaling pathways that can mediate G 2 arrest, only c-Abl-mediated signaling was activated and regulated the more prolonged MMRdependent G 2 arrest response. These data support the theory of a direct role for MMR signaling to provide time or induce cell death (41) at the G 2 checkpoint to correct or eliminate cells containing mutagenic DNA lesions induced by MNNG exposure.

MATERIALS AND METHODS
Chemicals, Reagents, and Cell Treatments-MNNG (Sigma) was dissolved in Me 2 SO as a 100 mM stock solution and stored at Ϫ20°C. 6-TG (Sigma) was dissolved in 0.1 N NaOH, and 1 mM stock solutions were stored at Ϫ20°C. STI571 (Gleevec TM , Novartis, East Hanover, NJ) was dissolved in water at 5 mM, and stock solutions were stored at Ϫ20°C. PD166326, a c-Abl inhibitor, was synthesized by us and dissolved at 10 mM in Me 2 SO. Exposure of cells to IR was performed using a 137 Cs irradiator (6). Cells were treated with ultraviolet (UVB) radiation at 50 J/m 2 .
Cell Culture-MMR-deficient human HCT116 (parental) colon cancer cells and an isogenic MMR-corrected HCT116 3-6 derivative (corrected for hMLH1 expression by microcell transfer of an extra chromosome 3) were provided by Dr. C. R. Boland (Baylor College, Dallas, TX). BRCA1 (breast cancerassociated gene 1) and WT HCC1937 corrected cells were obtained from Dr. K. Yamane (Case Western Reserve University, Cleveland, OH). hMLH1-deficient RKO cells were transfected with cytomegalovirus-driven full-length hMLH1 cDNA (obtained from Drs. A. Buermeyer and R. M. Liskay (Oregon Health Sciences University, Portland, OR), and clones were isolated in G418 (400 g/ml) by limiting dilution. Clone RKO7 expressed hMLH1 protein levels comparable with WT cells, whereas clone RKO6 contained the hMLH1 expression vector, expressed neo R without stable hMLH1 or hPMS2 protein expression, and served as a negative control. U2OS-derived stable cells that conditionally regulate FLAG-tagged ATR-WT or kinase-dead ATR (ATR-KD) levels by doxycycline were provided by Drs. P. Nghiem and S. L. Schreiber (Harvard University) (14). Normal human fibroblasts (N2936B) and fibroblasts (AT2052) from ataxia telangiectasia (AT) patients were obtained from the NIGMS Human Genetic Mutant Cell Repository (Camden, NJ). Human ATM-deficient fibroblast YZ5 and pEBS7 cells (provided by Dr. Yosef Shilo) were stably transfected with ATM-WT and an empty vector, respectively. Mouse MEF1-1 (Gadd45␣ ؉/؉ ) and MEF11-1 (Gadd45␣ Ϫ/Ϫ ) cells were provided by Dr. Al Fornace (Georgetown University, Washington, D. C). and grown as described (15). All cells were maintained in Dulbecco's modified Eagle's medium (Cambrex Bio Science, Walkersville, MD) with 10% fetal bovine serum (HyClone, Logan, UT) supplemented with penicillin (10 units/ ml) and streptomycin (10 units/ml) in a humidified 95% air and 5% CO 2 atmosphere. All cells used were mycoplasma-free.
MMR Status-Comparative genomics was used to evaluate the microsatellite instability (MSI) status in RKO6 versus RKO7 cell lines. Arrays manufactured by NimbleGen (Madison, WI) were hybridized with RNA-free DNA extracts of RKO6 and RKO7 cell lines in duplicate, along with a Promega total human standard. Each array contained seven copies of 53,735 unique probes, including WT, single-mismatch, double-mismatch, and deletion probes for all possible 1-mer through 6-mer repeats and their complements. This array also contained seven copies of poly(A) and poly(T) probes ranging from 29-mer to 47-mer in length to specifically measure MSI. Total poly(A)/ poly(T) amplification of the genome at ϳ3900 loci, dramatically more than the five poly(A) mononucleotide markers (BAT-25, BAT-26, NR-21, NR-24, and MONO-27), was used to measure MSI (16). The data were robust multiarray analysis (RMA)-normalized; specificity was confirmed using mismatch probes; and two replicates for each of a total of 126 poly(A) and 126 poly(T) probes were averaged. The relative total poly(A)/ poly(T) content was measured as a ratio of averages for RKO6 and RKO7 cell line genomes.
Transient siRNA Knockdown-Transient SMARTpool siRNA targeting ATR in MMR ϩ RKO7 cells or a scrambled control (SrcII) was purchased from and used as directed by Dharmacon (Lafayette, CO). ATR expression was monitored by Western blotting (supplemental Fig. 1B). After transfection (48 h) when ATR levels were significantly knocked down and scrambled shRNA-transfected cells expressed basal ATR levels, cells were treated with MNNG and analyzed for G 2 arrest and apoptosis.
Colony-forming Ability Assays-Survival was assessed by colony-forming ability using standard techniques, wherein colonies of Ͼ50 normal-appearing cells were counted (6). Isogenic cells were treated for 1 h with MNNG or continuously with 6-TG for 6 -7 days. Pretreatment with STI571 or other c-Abl inhibitors consisted of a 2-h pretreatment prior to MNNG (1 h) exposure.
Cell Cycle Analyses-Human cells were synchronized by growth to Ͼ95% confluence and maintained for 48 h in Dulbecco's modified Eagle's medium containing low serum (0.1% fetal bovine serum). To initiate synchrony, confluent cells were trypsinized and reseeded at 1:15 in 10% fetal bovine serumcontaining Dulbecco's modified Eagle's medium. MNNG was added 16 h after release, corresponding to a time prior to S phase initiation, and beyond the p53-mediated G 1 checkpoint (5). Pretreatment with the STI571 inhibitor occurred 2 h prior to MNNG exposure (10 M, 1 h). At various times post-treatment, cells were fixed and treated with RNase A (100 g/ml) and propidium iodide (50 g/ml in phosphate-buffered saline) overnight at 4°C. Cells were then analyzed on a Coulter Epics XL flow cytometer (Beckman Coulter, Inc., Fullerton, CA). MPM-2 staining was performed to distinguish G 2 and M phases of the cell cycle as described (5). Quantification of cell cycle populations was assessed using ModFit LT Version 3.0 software (Verity Software House, Topsham, ME). The results presented were from three or more independent experiments.
Nonradioactive c-Abl kinase assays were performed as described (SignaTECT, Promega) with modifications. Cell extracts were prepared as described (5) with addition of two phosphatase inhibitor mixtures (sc-45044 and sc-45045, Santa Cruz Biotechnology). GST-Abltide peptide (ϳ28 kDa; Upstate Biotechnology, Lake Placid, NY) was used as a high affinity c-Abl substrate; GST-Abltide contains the sequence EAIYAAPFAKKK, identical to GST-CRK. Kinase reactions were initiated with cell extracts (5 g) and Abltide substrate (1 g) for 30 min. Separate c-Abl kinase reactions using purified protein were performed as controls. Extracts were resolved and visualized by Western blotting, and membranes were probed using an anti-phosphotyrosine antibody (PY99, Santa Cruz Biotechnology), normalized by Ponceau S staining.
Statistics-All experiments described were performed in triplicate with representative images shown. Quantification of protein levels or c-Abl kinase activities was performed by scanning x-ray films and analyzing scans using NIH Image J software. Statistical significance was determined using paired Student's t tests.

Restoration of hMLH1 Expression and MMR Function in
MMR-deficient RKO Cells-RKO cells lack hMLH1 expression due to promoter hypermethylation (4). hMLH1 expression was restored by stably transfecting RKO cells with cytomegalovirus-driven hMLH1 cDNA. Two separate clones designated RKO6 (MMR Ϫ ), which remained deficient for hMLH1 expression, and RKO7 (MMR ϩ ), which expressed hMLH1 protein levels equivalent to MMR-proficient HCT116 3-6 cells (Fig. 1A) corrected by chromosome 3 microcell fusion (6, 7), were isolated and chosen for further investigation. RKO7 (MMR ϩ ) cells showed stable hPMS2 protein levels; hMLH1 and hPMS2 are binding partners (MutL␣) that require each other for stability. Notably, hMLH1 protein was not expressed in MMR-deficient RKO6 cells, as found in parental RKO cells. Using a compara-tive genomic hybridization array-based method, the MSI status of the RKO6 versus RKO7 genome was quantitatively measured at ϳ3900 poly(A)/poly(T) sites within the human genome. A Ͼ2.3 Ϯ 0.1-fold reduction of poly(A)/poly(T) content in the RKO7 genome relative to the RKO6 genome was observed, thus confirming its restoration to a more MSI-stable genome.
To determine whether re-expression of hMLH1 in RKO7 and RKO6 cells was functional, survival was assessed after MNNG or 6-TG exposure. RKO7 (MMR ϩ ) cells were signifi-cantly more sensitive than RKO6 cells to each agent (p Ͻ 0.001) (Fig. 1, B and C). RKO7 cells were as sensitive as HCT116 3-6 (MMR ϩ ) cells after MNNG exposure, whereas RKO6 cells were as resistant as hMLH1-deficient HCT116 P cells (Fig. 1B). RKO7 cells were also more sensitive than RKO6 (MMR Ϫ ) cells to 6-TG (p Ͻ 0.001) (Fig. 1C). These results demonstrated that restoration of MMR by exogenous hMLH1 expression in RKO7 cells enhanced their sensitivity to MNNG or 6-TG. In contrast, MMR-deficient RKO6 and HCT116 P cells were resistant and damage-tolerant.
Expression of hMLH1 Restores MMR-dependent G 2 Arrest Responses-To determine whether the RKO clones established above (Fig. 1A) were concomitantly corrected for MMR-dependent G 2 arrest responses (6, 7), we synchronized cells as described under "Materials and Methods," treated them with MNNG (5 M for HCT116 and 10 M for RKO, 1 h) at the indicated times (16 h after synchronous release) ( Fig. 2A), and examined their cell cycle profiles as described (6,7). In response to MNNG, both MMR ϩ and MMR Ϫ cells showed an initial G 2 arrest response 20 h after MNNG exposure. Thereafter, MMRdeficient RKO6 cells rapidly progressed through the cell cycle. In contrast, MMR-corrected RKO7 cells exhibited a prolonged G 2 arrest response (6,7). However, because these cells were synchronized, we noted that, contrary to other groups (17), MNNG-dependent G 2 arrest occurred within the first round of replication after treatment. Differential cell cycle G 2 arrest was also observed in matched RKO clones following continuous exposure to 2.5 M 6-TG (data not shown).
Changes in expression of cyclin B 1 and Cdc2 protein levels were examined to confirm differential G 2 arrest responses in MMR-proficient versus MMR-deficient cells. Elevated levels of cyclin B 1 , corresponding to the relative number of cells in G 2 , were observed in MNNG-treated MMR-competent cells (Fig.  2B). The activity of Cdc2, a G 2 cyclin-dependent kinase, depends on its association with cyclin B 1 and its phosphorylation of specific amino acid residues (18,19). In particular, the specific phosphorylation of Tyr 15 inhibits its activity, preventing cells from migrating through G 2 to M phase of the cell cycle. Using an antibody that specifically recognizes phosphorylated Tyr 15 of Cdc2, we noted enhanced levels of Cdc2 Tyr 15 phosphorylation in MMR-proficient cells, corresponding directly to higher levels of G 2 -arrested cells (Fig. 2B). In contrast, total cellular Cdc2 levels were constant throughout the cell cycle. ␤-Actin served as a loading control, with constant levels noted.
We originally noted that p53 was preferentially stabilized in MNNG-, 6-TG-, or IR-treated MMR-proficient cells, a response later confirmed by others (6,20). Using an antibody specific for p53 phosphorylated at Ser 15 , we showed that MMRproficient HCT116 3-6 and RKO7 cells displayed elevated p53 Ser 15 phosphorylation and stabilized p53 levels compared with isogenic MMR Ϫ cells (41). Interestingly, p21 (an inhibitor of cyclin-dependent kinase and a well characterized p53-dependent downstream gene) displayed similar induction levels irrespective of MMR status after MNNG treatment (41). Nevertheless, selective MMR-dependent p53 phosphorylation (and stabilization) was theorized to be a result of upstream signaling by phosphatidylinositol 3-like kinases such as ATM and its related kinase ATR, as well as their downstream checkpoint kinases Chk1 and Chk2, respectively. We therefore examined the potential involvement of the ATM/Chk2 and ATR/Chk1 signaling pathways in MMR-dependent G 2 arrest responses after MNNG exposure.
ATR can activate Chk1 and ATM can activate Chk2 by phosphorylating specific amino acid residues (21). Using antibodies that specifically recognize these phosphorylated forms of Chk1 and Chk2, we found that both RKO6 and RKO7 cells displayed equivalent Chk1 phosphorylation levels at Ser 296 , Ser 317 , and Ser 345 in response to MNNG (Fig. 2C). Notably, no phosphorylation of Chk2 at Thr 68 was observed in either cell line after MNNG exposure. Interestingly, parental HCT116 cells appeared to have much greater overall levels of Chk2 compared with MMR-corrected HCT116 3-6 cells. The overall protein levels of Chk1 and Chk2 remained constant throughout the cell cycle, with no increases in phosphorylation of either Chk1 or Chk2 in synchronized untreated control cells (data not shown). UV radiation (50 J/m 2 ) and ionizing radiation (10 grays (Gy)) were used as positive controls for ATR and ATM activation and associated Chk1 and Chk2 damage-induced phosphorylation, respectively (Fig. 2C). These data suggest that ATR/Chk1, but not ATM/Chk2, is activated after MNNG exposure in synchronized cells and, notably, that activation of the ATR/Chk1 pathway is independent of MMR function.
Neither ATM nor ATR Mediates MMR-dependent G 2 Arrest-To explore the role of ATM in MMRdependent G 2 arrest responses, we compared synchronized AT (pEBS7) or ATM-corrected (YZ5) AT fibroblast cells for their responses to IR (0.75 Gy) versus MNNG (10 M, 1 h). YZ5 (ATM ϩ ) cells showed significantly greater (p Ͻ 0.05) G 2 arrest compared with pEBS7 cells with IR treatment (0.75 Gy), consistent with prior reports of a role for ATM in G 2 arrest after IR (22). In contrast, when treated with MNNG (10 M, 1 h), pEBS7 (ATM Ϫ/Ϫ ) cells showed prolonged G 2 arrest responses compared with YZ5 (ATM ϩ ) cells (Fig. 3A), similar to results found using synchronized AT (AT2052) fibroblast versus normal human fibroblast (N2639B) cells after MNNG exposure (10 M, 1 h) (supplemental Fig. 1A). We also explored the role of ATR in G 2 arrest responses using U2OS cells containing doxycycline-inducible ATR-WT or ATR-KD (Fig. 3B). As with ATM deficiency (Fig. 3A), loss of functional ATR due to ATR-KD expression did not influence G 2 arrest responses compared with U2OS cells expressing ATR-WT after various MNNG doses (Fig. 3, C-E). These data are consistent with the lack of differential MMR-dependent downstream Chk1 and Chk2 phosphorylation responses (Fig. 2) and strongly suggest that neither ATM nor ATR influences G 2 arrest after MNNG exposure. In contrast, loss of functional ATR greatly affected IR-induced G 2 arrest responses, as well as G 2 arrest induced by UVB light (14). Results with ATR-KD were confirmed using siRNA oligomer knockdown of ATR, wherein no significant change in MNNG-induced G 2 arrest was noted in MMR ϩ RKO7 ATR siRNA cells (where ATR levels were knocked down by ϳ85%) compared with MMR ϩ RKO7 cells exposed to scrambled siRNA oligomers (supplemental Fig. 1, B and C).
c-Abl Kinase Is Required for MMR-dependent G 2 Arrest-Because MMR-dependent activation of c-Abl kinase after cispla- tin exposure was reported previously (23), we examined a role for its activation in MNNG-induced G 2 arrest. Pretreatment of isogenic MMR ϩ/Ϫ RKO cells with the c-Abl inhibitor STI571 (Gleevec TM ) (25 M, 2 h) prior to MNNG exposure dramatically abrogated subsequent G 2 arrest responses (Fig. 4A); The BCR-ABL inhibitor STI571 was designed to target the BCR-ABL fusion gene product or Philadelphia translocation that causes Ͼ90% of chronic myelogenous leukemia (24). Interestingly, STI571 did not abrogate the more transient G 2 arrest responses observed at 20 h in MMR-deficient cells (Fig. 4A).
We also performed these experiments by treating cells with 5 or 10 M STI571 for 16 -24 h using previously published procedures (supplemental Fig. 2, A and B) (25). Consistent with data in Fig. 4 (A and B), we found that, at these lower concentrations of STI571, MNNG-induced MMR-dependent G 2 /M checkpoint responses were abrogated. These data further support a role for MMR-activated c-Abl activity in these checkpoint responses. Because STI571 can inhibit other tyrosine kinases in addition to c-Abl, we tested another c-Abl/Src kinase inhibitor, PD166326, which was developed to overcome STI571 resistance (24). Similar to STI571, pretreatment of MMR ϩ cells with PD166326 (5 M, 2 h) preferentially abrogated MMR-dependent G 2 arrest after MNNG exposure. Treatment with either STI571 or PD166326 alone had no effect on the cell cycle. These findings showed that pretreatment with either STI571 or PD166326 abrogated MMR-dependent G 2 arrest and strongly suggest that c-Abl kinase is critical for MMR-dependent G 2 arrest signaling.
Although STI571 is more specific for c-Abl than for other tyrosine kinases in vivo, it may inhibit related enzymes at higher doses (e.g. platelet-derived growth factor, Src, and c-Kit) (24). Therefore, we also employed a genetic approach using stable shRNA to specifically knock down c-Abl protein levels and enzymatic activity. We generated isogenically matched MMR-proficient and MMR-deficient RKO cell lines expressing specific shRNA targeting c-Abl (shABL) or scrambled sequence (shSCR) (Fig. 5A). Notably, hMLH1 protein levels were not affected in RKO7-shABL or RKO7-shSCR clones (Fig. 5A). RKO6-shABL clone 2 and RKO7-shABL clone 27 had the greatest c-Abl knockdown, with Ͻ60% and Ͻ80% reduced protein levels, respectively. The c-Abl-deficient (shABL) and shSCRcontaining RKO7 cells were then examined for differences in MNNG-induced G 2 arrest (Fig. 5B). Further time course examination of all clones revealed that loss of c-Abl protein and activity resulted in loss of G 2 arrest responses (Fig. 5C). Collectively, these results suggest that c-Abl plays an important and specific role in MMR-dependent G 2 arrest responses to FIGURE 3. MMR-dependent G 2 arrest responses are not mediated by ATM or ATR. A, AT pEBS7 (ATM Ϫ/Ϫ ) or corrected YZ5 (ATM ϩ/ϩ ) human fibroblasts with identical doubling times were synchronized, exposed to IR (0.75 Gy) and MNNG (10 M, 1 h) at 16 h post-synchronization, and analyzed for G 2 arrest responses. UT, untreated. B, doxycycline (Dox) induction of FLAG-tagged ATR-WT or dominant-negative ATR-KD was confirmed in U2OS cells. NS, nonspecific band for loading. C-F, asynchronous cells from B were exposed to 1.5, 5.0, or 10 M MNNG for 1 h or to IR (10 Gy), respectively. G 2 arrest responses were monitored at various times. Changes in G 2 arrest were noted only after IR exposure (10 Gy). *, p Ͻ 0.05.
MNNG. Furthermore, c-Abl does not play a role in the more transient MMR-independent G 2 arrest responses observed, which are probably mediated by ATR/Chk1 signaling.
Stabilization of p73␣, Expression of GADD45␣, and Enhanced c-Abl Tyrosine Kinase Activity after MNNG Treatment Are MMR-dependent-We then sought direct evidence for MMR-dependent c-Abl activation. The p53 homolog p73␣ is a known downstream responsive gene controlled by c-Abl (26), and c-Abl-dependent stabilization of p73␣ after cisplatin exposure requires functional MMR (13). Stabilization during exposure to specific DNA-damaging agents is dependent on c-Abl-mediated phosphorylation at Tyr 99 of p73␣ (27). We examined p73␣ expression in MMR Ϫ RKO6 and MMR ϩ RKO7 cells after MNNG treatment (10 M, 1 h) by Western analyses (Fig. 6A). We noted a MMR-dependent increase in p73␣ expression 8 -56 h post-treatment in MMR ϩ RKO7 cells, but only a transient increase in p73␣ protein stabilization at 8 h in MNNG-treated RKO6 cells. Interestingly, MNNG-induced p73␣ protein stabilization mirrored G 2 arrest responses, with  increases prior to peak G 2 levels. Thus, p73␣ was preferentially stabilized in an MMR-dependent manner after MNNG exposure. Cisplatin (25 M, 24 h) and mock-exposed HCT116 3-6 cells were used as positive and negative controls, respectively, for p73␣ stabilization as described (13).
Prior work from our laboratory showed that GADD45␣ levels were elevated in MMR-proficient cells after FdUrd exposure (5). GADD45␣ was previously implicated in G 2 arrest regulation, making it a likely candidate in controlling MMR-dependent G 2 arrest (10). We therefore determined whether GADD45␣ expression was differentially expressed in isogenic MMR ϩ/Ϫ RKO cells after MNNG exposure. As we previously found after FdUrd exposure (5), MMR-dependent induction of GADD45␣ levels was noted after MNNG exposure (Fig. 6A). MMR-dependent GADD45␣ induction was also observed in isogenic HCT116 3-6 versus HCT116 cells after MNNG exposure (data not shown). Although minor transient increases in both p73␣ and GADD45␣ expression were also noted in MMRdeficient cells after MNNG treatment, these experiments clearly demonstrate the MMR dependence of expression for each of these proteins.
Because stabilization of p73␣ after cisplatin treatment is attributed to MMR-dependent increases in c-Abl tyrosine kinase activity (13), we wanted to determine whether c-Abl kinase activity was directly activated in an MMR-dependent manner after MNNG exposure. We therefore assessed the relative levels of c-Abl kinase activities by measuring phosphorylation at a specific recognition protein sequence within the GST-Abltide peptide. c-Abl activities were measured in synchronized MMR Ϫ RKO6 and MMR ϩ RKO7 cells before and after MNNG exposure. Enhanced GST-Abltide tyrosine phosphorylation in MMR-proficient cells treated with either cisplatin (25 M, 24 h) or MNNG (10 M, 1 h) was noted ( Fig. 6B; quantified in C). As with p73␣ stabilization, the levels of c-Abl kinase activity, judged by tyrosine phosphorylation, correlated with G 2 arrest responses in MMR ϩ RKO7 cells (Fig. 2). In contrast, MMR Ϫ RKO6 cells showed transient increases in c-Abl activity from basal levels at 8 h (Fig. 6, B and C), and these corresponded with transient increases in p73␣ and GADD45␣ protein expression (Fig. 6A) and G 2 arrest (Fig. 2).
GADD45␣, But Not p73␣, Regulates MMR-dependent G 2 Arrest-Because upstream MMR-dependent activation of c-Abl appears to regulate p73␣ and GADD45␣ levels, we explored the functional roles of p73␣ and GADD45␣ in MMRdependent G 2 arrest responses to MNNG. We generated knockdown cells using shRNA lentivirus specific for p73␣ (Fig.  8A) or GADD45␣ (shGADD45␣) (Fig. 8C). RKO6-shp73 clone 15 and RKO7-shp73 clone 4 had the greatest p73␣ knockdown, with 90% reduced protein levels each compared with shSCRcontaining RKO cells. Surprisingly, silencing p73␣ failed to abrogate MMR-dependent G 2 arrest despite its pro-survival effects on apoptosis and survival (Fig. 8B) (41). Thus, knocking down p73␣ in RKO7 cells significantly abrogated the apoptotic and lethality responses to MNNG without affecting MMR-dependent G 2 arrest. These data are the first to demonstrate uncoupling of the MMR-dependent G 2 arrest signaling from lethality.

Role of c-Abl in MMR-dependent G 2 Arrest Responses
Silencing GADD45␣ significantly abrogated MMR-dependent G 2 arrest responses to MNNG (p Ͻ 0.05) (Fig. 8D and supplemental Fig. 3), as well as apoptosis and survival (41). These data strongly suggest that MMR-dependent G 2 arrest responses to MNNG are regulated by GADD45␣ and not by p73␣.

DISCUSSION
MMR senses and repairs damaged lesions, but also appears to directly activate cell cycle arrest signaling pathways. The long-held theory that cells arrest at specific checkpoints, such as in G 2 , to allow time for repair of mutations prior to cell cycle  . c-Abl-mediated stabilization of GADD45␣, but not p73␣, regulates G 2 arrest responses in MMR ؉ cells after MNNG treatment. A, generation of p73␣ knockdown RKO6 and RKO7 clones. RKO6 (clones 11, 12, and 15) and RKO7 (clones 4, 12, and 18) clones expressing stable shp73 or a control shSCR vector were generated. Note that p73␤ levels were not affected in knockdown clones. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. B, specific knockdown of p73␣ does not abrogate G 2 arrest responses to MNNG (10 M, 1 h). UT, untreated. C, generation of GADD45␣ knockdown RKO6 and RKO7 clones. RKO6 (clones 1, 2, and 4) and RKO7 (clones 2, 3, and 4) clones expressing stable shGADD45␣ or a control shSCR vector were generated. D, RKO7-shSCR, RKO7-shGADD45␣-2, and RKO7-shGADD45␣-3 clones were exposed to MNNG (10 M, 1 h) and assessed for G 2 arrest. Specific knockdown of GADD45␣ abrogated G 2 arrest responses to MNNG. *, p Ͻ 0.05. progression appears to be very applicable to MMR processes. MMR-dependent G 2 arrest and elimination of severely damaged cells by apoptotic responses clearly result in mutational avoidance in human MMR ϩ cells that is several orders of magnitude lower than that in matched MMR Ϫ cells (28). To understand the signal transduction processes involved in these mutational avoidance processes, we functionally reconstituted MMR in hMLH1-deficient RKO cells. Stable hMLH1 re-expression in synchronized RKO cells restored MMR-dependent G 2 arrest and sensitivity to MNNG. For the first time, using both genetic and pharmacological approaches, we elucidated a role for c-Abl kinase in G 2 arrest responses in MMR ϩ cells after MNNG treatment. Disrupting c-Abl kinase (a) abrogated MMR-dependentG 2 arrest,(b)decreasedexpressionofMMRdependent GADD45␣ protein and p73␣ stabilization, and (c) abrogated MMR-dependent lethality (i.e. damage tolerance) in response to MNNG. Our data provide the first evidence that MMR-dependent G 2 arrest responses are dependent on c-Abl activation. Our results suggest that futile cycling may not be a means to MMR-dependent cell death because DNA breaks created by processing DNA lesions by MMR would be upstream from c-Abl/p73␣ activation. Abrogating c-Abl activity or p73␣ levels should not have affected MMR-mediated DNA single-strand breaks or DSBs and therefore lethality, as predicted (5,6). Abrogating c-Abl, GADD45␣, and, even more so, p73␣ leads to damage tolerance and increased survival, presumably without affecting MMR activity. These data suggest that if DNA single-strand breaks or DSBs are being created by MMR processing of damage, they do not mediate G 2 arrest and/or lethality. Thus, MMR appears to detect specific, potentially mutagenic MNNG-induced DNA lesions and selectively signals to c-Abl kinase to regulate G 2 arrest, apoptosis (41), and lethality. c-Abl activation causes GADD45␣ up-regulation, which we propose is a major determinant in MMR-dependent G 2 arrest, suggesting that MMR directly signals this checkpoint response, leading to cell death (Fig. 8) (41). Notably, inhibition of c-Abl by STI571 exposure or stable c-Abl shRNA knockdown resulted in increased MNNG resistance, raising the question of whether DSBs are formed because MMR processes are not affected. Preliminary assays using comet assays suggest that DSBs are formed via MMR-dependent processes, but are not affected by STI571 or c-Abl knockdown.
Although a link between MMR and c-Abl activation has been made for MMR-dependent apoptosis mediated by p73␣ after cisplatin exposure (13), the link to G 2 arrest has not. c-Abl kinase can phosphorylate Tyr 99 of p73␣ protein, stabilizing it in a manner similar to p53. Cisplatin also activates ATM, which, in turn, stimulates c-Abl, leading to p73␣ accumulation (13). p73␣ accumulation then directly stimulates apoptosis.
The mechanism by which MMR-dependent c-Abl mediates increases in GADD45␣ expression and how GADD45␣ controls both G 2 arrest and apoptosis remain unknown. MMR-dependent GADD45␣ protein expression was originally shown after FdUrd or 5-fluorouracil exposure (5). These responses occurred in the first cell division after drug exposure, wherein MMR directly detected fluoro-Ura:Gua mispairing (5). Overexpression of GADD45␣ can mediate G 2 arrest by direct binding and inhibiting Cdc2 (10). Loss of GADD45␣ leads to defective G 2 checkpoint signaling after DNA damage (10). Here, we established a link between GADD45␣ and MMR/c-Abl signaling. GADD45␣ can directly signal to p38␣ kinase (MAPK) after DNA damage and oncogenic transformation by Ras (29). We believe that this puts p38␣ stress kinase downstream from MMR/c-Abl/GADD45␣ signaling. p38␣ activity was previously linked to MMR-mediated G 2 arrest in response to temozolomide (an MNNG analog) (12). Thus, p38␣ may well be part of the c-Abl-mediated G 2 arrest pathway. Indeed, blocking G 2 arrest by p38␣ inhibition has no effect on Chk1 phosphorylation after temozolomide treatment (12), similar to our results after MNNG exposure (Fig. 2C). Indeed, both p38␣ and Chk1 FIGURE 9. c-Abl inhibition, by STI571 or genetic silencing, suppresses MMR-dependent lethality after MNNG treatment. A, survival assays of isogenic MMR ϩ and MMR Ϫ RKO cells pretreated with STI571 (25 M, 2 h) and exposed to varying doses of MNNG (M, 1 h). B, survival of isogenic MMR ϩ and MMR Ϫ RKO-shSCR clones compared with c-Abl knockdown shABL clones exposed to various MNNG doses (M, 1 h). C, survival of MMR ϩ RKO7 and MMR ϩ RKO7-shABL knockdown clones pretreated with various STI571 doses (M, 2 h) prior to MNNG exposure (10 M, 1 h). signal transduction pathways contribute separately, but equally, to MMR-dependent temozolomide-induced G 2 arrest (30). Thus, we theorize that p38␣ kinase is downstream not only from c-Abl, but also from GADD45␣. More direct evidence for involvement of GADD45␣ in MMR-dependent lethality is shown in the accompanying article (41).
Early onset BRCA1 is a potential candidate protein that could bridge the gap between MMR and c-Abl signaling to control GADD45␣ induction. BRCA1 has been proposed to play a dual role in G 2 arrest and homologous DSB repair. BRCA1 can be phosphorylated and activated after genotoxic stress, allowing it to act as a transcription factor for several downstream genes, including GADD45␣. Thus, MMR-dependent GADD45␣ induction (Fig. 6A) may be a result of BRCA1 activation, linked to c-Abl activity in response to MMR signaling. Evidence of a role for BRCA1 in MMR-dependent signaling, independent of ATR or Chk1 for lethality, was reported recently (31). BRCA1 can be activated by ATM kinase after a DSB, disrupting its interaction from a stable inactive complex with c-Abl (32). This allows both proteins to separate and become active. A direct interaction between MMR and BRCA1 was noted (33), and BRCA1 can physically interact with hMSH2 and two of its binding partners, hMSH3 and hMSH6 (34). BRCA1 and hMSH2 are part of a proposed multiprotein complex involved in DNA damage recognition and repair, known as BASC (BRCA1-associated genome surveillance complex) (35). Recently, Yamane et al. (31) noted a minor role for BRCA1 in MMR-dependent G 2 arrest signaling, similar to our own findings (supplemental Fig. 4); however, caution should still be taken in attributing a role for BRCA1 in MMR-specific processes without more detailed investigations.
Recently, Kim et al. (25) found that c-Abl may physically interact with hMLH1; however, they failed to evaluate how this interaction might affect cell cycle and apoptotic responses. They also determined that MMR activates the JNK/MAPK signaling cascade after MNNG exposure; however, these results are not surprising because MMR-dependent activation of the c-Abl/JNK pathway has previously been observed after cisplatin exposure (36). A direct interaction between c-Abl and hMSH5 has also been shown (37). In fact, a single point mutation (Y823H) in hMSH5, acting similarly to loss of hMSH2, results in damage tolerance to MNNG or temozolomide (38). Because the hMSH4-hMSH5 heterodimer is involved in Holliday junction recognition and resolution (39), hMSH4-hMSH5 may stabilize and preserve a meiotic bimolecular DSB repair intermediate. How MMR-specific c-Abl kinase activation participates in these processes will require further research.
The dramatic pro-survival effects achieved by inhibiting c-Abl kinase activity using STI571 or c-Abl kinase silencing on MNNG-dependent cytotoxicity have major clinical implications. Others have demonstrated that c-Abl inhibition by Gleevec TM can protect cells from DNA-damaging agents, including H 2 O 2 , Ara-C, and curcumin (40). Collectively, these results suggest that caution should be taken in clinical protocols that combine STI571 (or other c-Abl/Src inhibitors) with chemotherapeutic agents that are modulated by MMR (e.g. cisplatin, 5-fluorouracil, and temozolomide).