DNA Mismatch Repair-dependent Activation of c-Abl/p73α/GADD45α-mediated Apoptosis*

Cells with functional DNA mismatch repair (MMR) stimulate G2 cell cycle checkpoint arrest and apoptosis in response to N-methyl-N′-nitro-N-nitrosoguanidine (MNNG). MMR-deficient cells fail to detect MNNG-induced DNA damage, resulting in the survival of “mutator” cells. The retrograde (nucleus-to-cytoplasm) signaling that initiates MMR-dependent G2 arrest and cell death remains undefined. Since MMR-dependent phosphorylation and stabilization of p53 were noted, we investigated its role(s) in G2 arrest and apoptosis. Loss of p53 function by E6 expression, dominant-negative p53, or stable p53 knockdown failed to prevent MMR-dependent G2 arrest, apoptosis, or lethality. MMR-dependent c-Abl-mediated p73α and GADD45α protein up-regulation after MNNG exposure prompted us to examine c-Abl/p73α/GADD45α signaling in cell death responses. STI571 (Gleevec™, a c-Abl tyrosine kinase inhibitor) and stable c-Abl, p73α, and GADD45α knockdown prevented MMR-dependent apoptosis. Interestingly, stable p73α knockdown blocked MMR-dependent apoptosis, but not G2 arrest, thereby uncoupling G2 arrest from lethality. Thus, MMR-dependent intrinsic apoptosis is p53-independent, but stimulated by hMLH1/c-Abl/p73α/GADD45α retrograde signaling.

cesses, causing cytochrome c release and caspase-9 activation (11). Reciprocally, once caspase-9 is activated, it can stimulate caspase-8 activation. Thus, the temporal kinetics of initial caspase-9 versus caspase-8 activation can be used to differentiate intrinsic versus extrinsic apoptotic signaling pathways, respectively.
Alkylating agents are commonly used cancer chemotherapeutic agents that can kill cells by apoptotic processes (12). Monofunctional alkylating agents can be divided into two major groups, S n 1 and S n 2. S n 2-type alkylating agents include methyl methanesulfonate and dimethyl sulfate. S n 1-type alkylating agents include temozolomide, MNNG, and N-methyl-NЈ-nitrosourea and are cytotoxic to cancer cells at lower concentrations than S n 2-type agents. S n 1-type compounds have therefore been used more often as antitumor agents with limited success. Treatment of cancer cells with S n 1-type agents such as MNNG and temozolomide gives rise to N 7 -methylguanine, N 3 -methyladenine, O 4 -methylthymine, and O 6 -methylguanine DNA lesions. N 7 -Methylguanine and N 3 -methyladenine DNA lesions are removed by base excision repair. O 6 -Methylguanine can be repaired by O 6 -methylguanine-DNA methyltransferase, which removes the methyl group from guanine in a single step and transfers it to an internal cysteine residue (Cys 145 ) on O 6 -methylguanine-DNA methyltransferase (as reviewed in Ref. 13). This reaction inactivates O 6 -methylguanine-DNA methyltransferase, but directly restores guanine in DNA without DNA incision, synthesis, or ligation steps. Cells that do not express detectable O 6 -methylguanine-DNA methyltransferase levels or are treated with O 6 -benzylguanine (a selective O 6 -methylguanine-DNA methyltransferase inhibitor) are hypersensitive to O 6 -methylguanine-inducing agents (13). Other DNA repair pathways such as nucleotide excision repair and MMR may also process O 6 -methylguanine lesions (4). A hallmark of MMR-mediated response to these DNA lesions is prolonged G 2 arrest and lethality. In contrast, MMRdeficient cells are highlighted by "damage tolerance" with heightened mutation rates (14). Defining the exact retrograde signaling events from MMR-specific DNA lesion detection to G 2 arrest and apoptosis remains important in understanding the efficiency of mutational avoidance of the MMR system.
Numerous studies have attempted to elucidate the signaling events that occur between MMR detection of DNA lesions in the nucleus and those involved in programmed cell death. Alkylation-induced signaling and apoptotic responses require MutS␣ (the hMSH2-hMSH6 heterodimer) and MutL␣ (the hMLH1-hPMS2 heterodimer) protein complexes (9,15) and human exonuclease I (16). A significant number of signaling molecules, including p53, Chk1, Chk2, ATR, ATM, and p38␣ mitogen-activated protein kinase (MAPK), were reported to regulate MMR-dependent signaling cascades, culminating in G 2 arrest and apoptosis (15,17,18). A recent study suggested that ATR, and not ATM, is activated in the presence of MutS␣ and MutL␣ protein complexes (19), although this evidence was derived from cell-free extracts.
In addition to the above-mentioned signaling pathways, a role for p53 in MMR-dependent apoptosis was also suggested. Our laboratory was the first to demonstrate a differential MMRdependent p53 stabilization after 6-thioguanine (6-TG) or IR exposure (14). Results from other laboratories suggested that MMR-dependent apoptosis in human colon cancer cells after MNNG or N-methyl-NЈ-nitrosourea treatment is dependent on p53 and hMLH1, although isogenic cells were not used (20). Furthermore, temozolomide-induced apoptotic responses were reported to be dependent on p53 (21), although these responses were not examined for MMR dependence. In contrast, MNNG-or temozolomide-induced apoptosis was observed in TK6 cells that express viral E6 protein, making cells functionally null for p53 (22). Other studies suggested that p73, a homolog of p53, is activated in response to cisplatin in an MMR-specific manner (23)(24)(25). Indeed, hPMS2 (a binding partner of hMLH1 and component of the hMutL␣ complex) may directly interact with stabilized p73␣ after cisplatin exposure (26). A role for p73␣ in MMR-dependent apoptosis was not, however, mechanistically explored. Thus, to date, an indepth study of MMR-dependent retrograde (nucleus-to-cytoplasm) signaling that culminates in cell death (apoptosis) has not been performed.
Using isogenic hMLH1-deficient versus hMLH1-corrected human colon cancer cells, we elucidated the hMLH1/c-Abl/ p73␣/GADD45␣ (growth arrest-and DNA damage-inducible-45␣) signaling pathway that controls apoptosis in response to MNNG treatment. We demonstrate functional roles for c-Abl and GADD45␣ in MMR-dependent G 2 arrest, apoptosis, and lethality. We also demonstrate a specific role for c-Abl-induced p73␣ stabilization in apoptotic responses and survival after MNNG treatment, although this p53 homolog plays no role in G 2 arrest responses. Thus, we provide the first evidence that MMR-dependent G 2 arrest responses can be uncoupled from apoptosis initiation. G 2 arrest may therefore be required for mutational avoidance, but not for apoptosis or lethality. Finally, we show that although p53 is differentially stabilized in an MMR-dependent manner, this signaling response plays no apparent role(s) in G 2 arrest responses, apoptosis, or lethality after MNNG treatment. Clinically, our data strongly suggest that inhibitors of c-Abl such as Gleevec TM are ill suited for combination therapy with temozolomide and cisplatin, which depend on functional MMR responses for efficacy.

MATERIALS AND METHODS
Reagents and Chemicals-MNNG, staurosporine, propidium iodide, puromycin, and RNase A were purchased from the Sigma. MNNG was dissolved in Me 2 SO at 100 mM and stored at Ϫ20°C. STI571 (Gleevec TM , Novartis, East Hanover, NJ) was dissolved in water at 5 mM and stored at Ϫ20°C. Stock concentrations were determined by spectrophotometric analyses.
Plasmids and Short Hairpin RNA (shRNA)-Full-length hMLH1 cDNA was generously provided by Drs. A. Buermeyer and R. M. Liskay (Oregon Health Sciences University, Portland, OR). Human DD1 transcriptionally inactive, dominant-negative p53 mutant cDNA was a kind gift from Dr. George Stark (Cleveland Clinic, Cleveland, OH). Human p73␣ and p73␤ isoforms were kind gifts from Dr. Meredith Irwin (Hospital for Sick Children Research Institute, Ontario, Canada). Human papillomavirus E6 cDNA was obtained from Dr. C. Reznick (University of Wisconsin Medical School, Madison, WI), cloned into a mammalian cytomegalovirus expression vector, and used to transfect RKO or HCT116 cells as indicated (14). Human scrambled shRNA (shSCR) sequence and p53 shRNA (shp53) were obtained from Dr. M. Jackson (Case Western Reserve University, Cleveland) (27).
All drug treatments were performed in medium lacking antibiotics or selective agents. For alkylation exposure, cells were incubated with MNNG (2-10 M, 1 h). Where indicated, cells were pretreated with 25 M STI571 for 2 h prior to MNNG exposure. Cells were then washed and scraped, and whole cell extracts were harvested at 24 -96 h as indicated. Cells were also pretreated as indicated with the caspase inhibitor Z-VAD-fmk (caspase inhibitor I, EMD Biosciences, La Jolla, CA) or Z-DEVD-fmk (caspase-3 inhibitor II, EMD Biosciences) at 20 or 50 M, respectively. Colony-forming assays were then performed (14) three or more times in triplicate, and statistics were completed as described below.
Flow Cytometry-For cell cycle distribution assays, adherent and floating cells were treated with 50 g/ml propidium iodide and 100 g/ml RNase A at 4°C and analyzed for DNA content and apoptotic populations using TUNEL assays (29). Cells were then analyzed using a FACSCalibur flow cytometer (BD Biosciences), and 10,000 events were plotted using CellQuest software.
Statistics-All studies were performed in triplicate at a minimum. The Western blots presented are data from experiments performed three times with similar results. Quantification of protein levels was performed by scanning x-ray films and analyzing the scans using NIH Image J software. Statistical significance was determined using paired Student's t tests.

RESULTS
Lethality Correlates with Apoptosis in MNNG-treated MMRcorrected RKO Cells-Human RKO cells are MMR-deficient due to hMLH1 promoter hypermethylation and are incapable of forming the hMutL␣ complex (i.e. hMLH1-hPMS2 heterodimer), rendering them deficient in both hMLH1 and hPMS2 protein expression. A prior study reported that hMLH1 overexpression is toxic (31); however, we selected RKO clones that express hMLH1 at levels comparable to endogenous protein amounts in wild-type cells. Selection of clones with hMLH1 levels comparable to endogenous levels in wild-type cells was critical for long-term stability. Various RKO clones were established, and all showed increased sensitivity to MNNG or 6-TG ( Fig. 1) (46). However, no clone demonstrated polyploidy in 3 or more years in culture, unlike previous reports of RKO clones in which hMLH1 expression was driven to abnormally high levels (32).
MMR ϩ clone RKO7 was chosen to further explore the role of hMLH1 expression in apoptotic responses after MNNG treatment. A dramatic but delayed MMR-dependent apoptotic response in RKO7 cells after a single dose of MNNG (10 M, 1 h) was observed, with peak levels appearing at 72-96 h postexposure (Fig. 1A). Apoptosis gradually declined in RKO7 cells at times Ͼ96 h, until few cells remained. MMR ϩ RKO7 cells showed a strong correlation between lethality (i.e. loss of colony-forming ability) and apoptosis induction at 96 h (Fig. 1B). In contrast, MMR Ϫ RKO6 cells were not responsive to MNNG in terms of apoptosis (Fig. 1B), even at doses Ն10 M, where loss of survival was noted. In fact, RKO cells exposed to 20 M MNNG did not elicit appreciable apoptotic responses, but did show an enlarged and "flattened out" morphology. Cell death in MMR Ϫ RKO6 cells after high dose MNNG exposure was not caused by apoptosis. To determine that this MMR-dependent apoptosis response was unique to MNNG treatment, we treated MMR-proficient and MMR-deficient cells with cisplatin. Using TUNEL as an output for apoptosis, we found that there was no appreciable difference in the amount of apoptosis measured in MMR-proficient and MMR-deficient cells (supplemental Fig. 1).

MMR-dependent Apoptosis Induced by MNNG Occurs via
the Intrinsic Pathway-To determine whether MMRdependent apoptosis in response to MNNG (10 M, 1 h) occurred through intrinsic or extrinsic apoptotic pathways, we monitored the kinetics of caspase-8 and caspase-9 as well as PARP-1 proteolyses in RKO6 (MMR Ϫ ) versus RKO7 (MMR ϩ ) cells. Significant caspase-9 activation cleavage (35 kDa) occurred at 48 h post-treatment in RKO7 cells at a time concomitant with significant levels of apoptosis ( Fig. 2A,  arrow). Notably, caspase-9 cleavage was noted at 48 h posttreatment, nearly 20 h prior to the activation cleavage of caspase-8 (43 kDa; arrow) observed at 72 h. Activation of caspase-9 prior to caspase-8 strongly suggested that the intrinsic apoptotic pathway was activated by MMR in response to MNNG. Pretreatment of MNNG-treated RKO7 cells with the caspase inhibitor Z-VAD-fmk or Z-DEVD-fmk prevented PARP-1 cleavage and abrogated apoptosis (Fig.   2B), suggesting that MMR-mediated apoptosis is caspasedependent. In contrast, RKO6 (MMR Ϫ ) cells did not undergo apoptosis in response to MNNG. Exposure of MMR ϩ and MMR Ϫ cells to staurosporine caused apoptosis equally in both cells independent of MMR function ( Fig. 2A). Thus, RKO6 (MMR Ϫ ) cells are functionally capable of inducing apoptosis in response to staurosporine and do not lack the capacity to die by this mechanism.
Abrogating p53 Function Does Not Affect MMR-dependent Apoptosis-Our laboratory was the first to report differential MMR-dependent stabilization of p53 in cells after 6-TG or IR treatment, and we hypothesized that MMR-dependent cell death and lethality may be p53-dependent (14). We confirmed differential p53 stabilization and phosphorylation (Ser 15 ) in RKO7 (MMR ϩ ) versus RKO6 (MMR Ϫ ) cells in response to MNNG (Fig. 3A), a signaling pathway that would be consistent with the activation of ATM, ATR, or DNA-dependent protein kinase (33)(34)(35)(36). Furthermore, significant increases in Bax and p21 expression, known downstream p53 target genes, were observed in RKO7 cells in a prolonged manner compared with transient responses in RKO6 cells (Fig. 3A). As with G 2 arrest responses (see Fig. 2 in the accompanying article) (46), RKO6 cells responded initially to MNNG by stabilizing and phosphorylating p53 (Ser 15 ) from 4 -12 h, with accompanying minor increases in p21 and Bax; however, the responses rapidly declined as noted (14). Thus, elevated transcriptionally functional p53 levels may be involved in MMR-dependent apoptosis after MNNG exposure. To examine its role in MMR signaling, we functionally FIGURE 1. Enhanced apoptosis in MMR-proficient cells after MNNG exposure. A, stable RKO6 (hMLH1 Ϫ /MMR Ϫ ) and genetically matched, hMLH1transfected RKO7 (hMLH1 ϩ /MMR ϩ ) cells were generated and analyzed (46). MMR Ϫ RKO6 versus MMR ϩ RKO7 cells were then exposed to MNNG (10 M, 1 h) and monitored for apoptosis using TUNEL and G 1 subpopulations. Apoptosis was dramatically increased in RKO7 (MMR ϩ ) cells versus RKO6 (MMR Ϫ ) cells at 48 h. B, RKO6 (MMR Ϫ ) and RKO7 (MMR ϩ ) cells were treated with increasing MNNG doses (M, 1 h), and apoptosis and survival end points were analyzed. *, p Ͻ 0.05; **, p Ͻ 0.01.
To further confirm that p53 was not required for MMRmediated apoptosis, we generated stable knocked down p53 using shp53 (Fig. 3C). Stable clones showed no detectable p53 protein levels in either RKO6 or RKO7 cells after stable infection with the shp53 lentiviral vector versus cells infected with the shSCR lentiviral vector (Fig. 3C). When RKO7-shp53 cells were treated with MNNG (10 M, 1 h), apoptosis occurred at 72 and 96 h with similar kinetics, and the extent of cell death observed was not statistically different from apoptosis in MNNG-treated RKO7-shSCR cells. Low levels of apoptosis were observed in both RKO6-shp53 and RKO6-shSCR cells (Fig. 3C), with no statistical differences noted. The same results were also observed upon stable expression of dominant-negative p53 (also noted as DD1) clones ( Fig. 2B and supplemental  Fig. 3).
To confirm loss of functional p53 in the cells described above, we treated human papillomavirus E6-, shp53-, and dominant-negative p53-expressing stable RKO cells with IR (8 Gy) and noted significant loss of G 1 arrest responses with respect to vector alone or shSCR counterparts (Fig. 3 and supplemental  Fig. 4). Loss of functional p53 in cells was also accompanied by subsequent loss of MNNG-induced Bax expression (Fig. 3 and  supplemental Fig. 5). Finally, loss of p53 function in MMR ϩ HCT116 or RKO cells did not affect the survival of cells exposed to various MNNG doses (see Fig. 5, A and B, and supplemental Fig. 6) or MNNG-induced MMR-dependent G 2 arrest responses in asynchronous or synchronized cells (supplemental Fig. 5, C and D). Collectively, these data strongly suggest that MMR-dependent cell death and G 2 arrest responses are not mediated by p53.
Enhanced p73␣ Stabilization in MMR ϩ Cells after MNNG Treatment-A role for p73 in stimulating apoptosis after cisplatin exposure was reported (23)(24)(25). We therefore examined p73 expression in RKO7 (MMR ϩ ) versus RKO6 (MMR Ϫ ) cells before and after MNNG exposure (10 M, 1 h). Increases in p73␣ (but not p73␤) expression were noted at 24 -48 h in both RKO7 (MMR ϩ ) and RKO6 (MMR Ϫ ) cells after MNNG treatment (Fig. 4, A and B). However, in a response apparently concomitant with p53 stabilization and G 2 arrest responses, p73␣ (but not p73␤) levels were elevated to a greater extent and in a more prolonged manner (at least to 72 h) (Fig. 4, A and B) in MMR ϩ RKO7 compared with MMR Ϫ RKO6 cells. Interestingly, we noted the consistent loss of p73␣ and p73␤ in RKO6 cells over time after MNNG treatment. Loss of p73 isoforms was consistent with a lack of apoptosis and may be indicative of loss of c-Abl activity (see below), growth arrest, or cell death (senescence) in RKO6 cells after MNNG treatment.
Given the role of c-Abl in MMR-dependent G 2 arrest and lethality (46), we examined the effect of STI571 (Gleevec TM ), a c-Abl tyrosine kinase inhibitor, on MMR-dependent p73␣ stabilization and apoptosis. Interestingly, MNNG-induced MMRdependent p73␣ increases, as well as basal p73␤ expression, were blocked by STI571 (Fig. 4C). Extracts from SK-N-AS cells that lack p73 expression and from cells separately expressing p73␣ or p73␤ were used to define the locations of p73 isoforms (Fig. 4, A and C). Collectively, these data strongly suggest that p73␣ expression is MMR-dependent and regulated by c-Abl activation after MNNG exposure.

DISCUSSION
The major goal of this and our accompanying study (46) was to elucidate the mechanism by which retrograde MMRdependent signaling induces G 2 arrest and apoptosis in response to MNNG-induced DNA lesions. In light of prior reports using cisplatin (23)(24)(25), our studies using MNNG strongly suggest a common signaling pathway involving hMLH1/c-Abl/GADD45␣ for G 2 arrest and p73␣ activation and stabilization for apoptosis (Fig. 8). Our data strongly suggest that 6-TG responses involve identical signaling pathways (data not shown), rejecting the theory that signaling from different lesions detected by MMR might be agent-specific.
Using genetic and pharmacological approaches, we have shown that exposure of cells to MNNG induced MMR-dependent (i) apoptosis via intrinsic apoptotic signaling, (ii) cell death processes that are not dependent on functional p53, and (iii) apoptosis signaled through the hMLH1/c-Abl/p73␣ or hMLH1/c-Abl/GADD45␣ pathway. Our data supply the first evidence that MMR-dependent apoptosis and G 2 arrest can be uncoupled. Down-regulation of either c-Abl or GADD45␣ blocked both apoptosis and G 2 arrest stimulated by MMR, whereas p73␣ knockdown affected only apoptosis, suggesting that MMR specifically detects MNNG-induced DNA lesions and signals directly to c-Abl, as confirmed in the accompanying article (46). c-Abl activation appears to be downstream from hMLH1 but upstream from p73␣ and GADD45␣ expression because shRNA-mediated knockdown of this tyrosine kinase or addition of STI571 (Gleevec TM ) dramatically decreased expression of GADD45␣ and p73␣, two proteins whose activities appear to regulate apoptosis as well as cell lethality.
For the first time, we have shown that MMR-dependent apoptosis occurs by delayed activities (after 48 h) of the intrinsic pathway. Caspase-9 was activated significantly earlier (ϳ20 h) than caspase-8. Caspase-9 is specifically activated by mitochondrial catastrophe, release of cytochrome c, and engagement of Apaf-1 (10). Its activation through hMLH1/c-Abl/p73␣/ GADD45␣ signaling must ultimately lead to expression of proapoptotic factors that mediate mitochondrial permeability transition.
Surprisingly, p53 stabilization and function are irrelevant to G 2 arrest, apoptosis, or survival in RKO or HCT116 cells (Fig. 3 and supplemental Figs. [2][3][4][5]. We anticipated that MMR-dependent stabilization of p53 (Fig. 3A) (14) would mediate apoptosis. Our data definitely show that p53 does not play a functional role in any MMR-dependent signaling responses triggered by detection of MNNG-induced DNA lesions. The p53 tumor suppressor has been implicated in several critical damage-induced pathways, including p21-mediated G 1 arrest, Bax induction, and pro-apoptotic protein expression. p53 induction also mediates GADD45␣-induced DNA repair and p21-and/or p16-mediated cellular senescence (37)(38)(39). Interestingly, we concomitantly noted p21 and Bax protein accumulation in MNNG-treated MMR ϩ cells, consistent with downstream transactivation by p53. It is surprising that such responses play no role in cell death because a major function of Bax and other known downstream pro-apoptotic p53-responsive genes (e.g. PUMA) are reported to translocate to mitochondria and to mediate cytochrome c and Apaf-1 release, activating Apaf-1/caspase-9 (40). MMR-dependent MNNG-induced G 2 arrest was not affected by abrogation of ATM, ATR, Chk1, or Chk2. Thus, detection of MNNG-induced DNA lesions by MMR leads to phosphorylation, stabilization, and transactivation of p53; however, its functions are not required for apoptosis and G 2 arrest. Similar conclusions were drawn using 6-TG or IR (14).
The exact mechanism and functional significance of MMRdependent phosphorylation and stabilization of p53 after MNNG exposure are unknown. We suspect that prolonged p53 response in MMR ϩ cells plays a major role in the control of subsequent G 1 arrest. Cells that escape G 2 arrest after MNNG exposure would be halted in G 1 before proceeding through another round of DNA synthesis, avoiding further amplification of mutations. DNA damage signaling is a major mechanism by which p53 becomes functionally activated (41). The exact mechanism by which p53 becomes activated by hMLH1/ c-Abl signaling remains unknown. The majority of DNA lesion sensors in the cell (i.e. DNA-dependent protein kinase, ATM, and ATR) are serine/threonine kinases that target p53 for phosphorylation and act to promote tetramerization (34,36). The mechanisms by which c-Abl might cross-talk through such enzymes is currently being explored.