Selenium Compounds Activate ATM-dependent DNA Damage Response via the Mismatch Repair Protein hMLH1 in Colorectal Cancer Cells*

Epidemiological and animal studies indicate that selenium supplementation suppresses risk of colorectal and other cancers. The majority of colorectal cancers are characterized by a defective DNA mismatch repair (MMR). Here, we have employed the MMR-deficient HCT 116 colorectal cancer cells and the MMR-proficient HCT 116 cells with hMLH1 complementation to investigate the role of hMLH1 in selenium-induced DNA damage response, a tumorigenesis barrier. The ATM (ataxia telangiectasia mutated) protein responds to clastogens and initiates DNA damage response. We show that hMLH1 complementation sensitizes HCT 116 cells to methylseleninic acid, methylselenocysteine, and sodium selenite via reactive oxygen species and facilitates the selenium-induced oxidative 8-oxoguanine damage, DNA breaks, G2/M checkpoint response, and ATM pathway activation. Pretreatment of the hMLH1-complemented HCT 116 cells with the antioxidant N-acetylcysteine or 2,2,6,6-tetramethylpiperidine-1-oxyl or the ATM kinase inhibitor KU55933 suppresses hMLH1-dependent DNA damage response to selenium exposure. Selenium treatment stimulates the association between hMLH1 and hPMS2 proteins, a heterodimer critical for functional MMR, in a manner dependent on ATM and reactive oxygen species. Taken together, the results suggest a new role of selenium in mitigating tumorigenesis by targeting the MMR pathway, whereby the lack of hMLH1 renders the HCT 116 colorectal cancer cells resistant to selenium-induced DNA damage response.

Selenium is an essential nutrient widely distributed in organic forms in certain foods and inorganic forms in soil. Selenium exerts its anticarcinogenic effects mainly through selenoproteins at nutritional levels and through selenium metabolites, including reactive oxygen species (ROS), 3 at supranutritional levels (1,2). Animal and epidemiological studies strongly implicate selenium as an effective chemoprevention agent against colon cancer (3)(4)(5)(6); however, the Nutritional Prevention of Cancer Trial and the recent Selenium and Vitamin E Cancer Prevention Trial reported mixed results on the efficacy of selenium in suppressing prostate cancer (7)(8)(9). Whatever the reason, the molecular mechanism by which selenium mitigates colorectal tumorigenesis is largely unknown.
The majority of colorectal cancers are characterized by microsatellite instability due to a defective MMR system (10,11). The MMR system senses DNA base mismatch after DNA replication and provokes repair, checkpoint and apoptotic responses (12). Among the many human MMR proteins, hMLH1 (human MutL homologue-1) and hPMS2 (human post-meiotic segregation protein-2) form a complex that recognizes and stabilizes mismatched DNA at an early stage and facilitates the DNA damage response (13). Somatic mutations in MMR genes and epigenetic silencing of hMLH1 expression are observed in a significant portion of sporadic colorectal cancers (8,14), whereas germ line mutations in hMLH1 account for 60% of the autosomal dominant nonpolyposis colon cancer. Loss of both hMLH1 and hMSH2 (human MutS homologue-2) is associated with complete inactivation of MMR, whereas defects in other MMR proteins result in only partial MMR deficiency (15).
The G 2 /M checkpoint prevents damaged cells from entering mitosis by coordinating DNA repair and/or apoptosis pathways. In response to DNA damage, hMLH1 can regulate G 2 /M transition through Cdc2 protein (16), and ATM kinase is implicated in G 2 /M checkpoint response (17)(18)(19). Moreover, a linkage between hMLH1 and ATM after DNA damage has been demonstrated (20,21). We have shown recently that selenium compounds at doses Յ LD 50 can activate DNA damage response through ATM and ROS in noncancerous but not in cancerous cells (22). In the current study, we hypothesized that lack of a G 2 /M checkpoint response renders the hMLH1-deficient HCT 116 cells resistant to selenium-induced DNA damage response. By employing isogenic cell lines with or without hMLH1 expression, we show herein that hMLH1 is required for
Immunoprecipitation and Immunoblotting Analysis-Immunoprecipitation and immunoblotting analyses were performed as described previously (26,27). Briefly, protein Gagarose-precleared lysates were immunoprecipitated with rabbit anti-PMS2 polyclonal antibodies (1:500, C-20, Santa Cruz Biotechnology, Santa Cruz, CA) for 16 h at 4°C. After washing three times in lysis buffer and after heat denaturation, proteins in the anti-PMS2 immunoprecipitates were separated by SDS-PAGE and transferred to a PVDF membrane, followed by sequential incubation with rabbit anti-MLH1 polyclonal antibodies (1:1000, C-20, Santa Cruz Biotechnology), HRPconjugated secondary antibodies, and the chemiluminescent reagents (SuperSignal, Pierce). Membranes were stripped using the Restore Western blot stripping buffer (Pierce) when neces-sary. ImageJ 1.43i (National Institutes of Health) was used to quantify the immunoblot signals on exposed films.
Flow Cytometric Assay-Cell cycle analysis was performed as described previously (22) with modifications. Briefly, HCT 116 and HCT 116ϩhMLH1 cells were seeded into 10-cm dishes and treated with selenium compounds for 0 -48 h, with or without pretreatment with KU55933 (10 M), NAC (10 mM), or Tempo (2 mM) for 24 h. Because G 2 /M cells tend to detach from the culture dishes, cell culture medium from selenium-treated cells were collected in an attempt to recover the loosely attached cells. Phases of the cell cycle were determined by flow cytometry (FACSCalibur cytometer with CELLQuest Pro, Becton Dickson, San Jose, CA) along with the ModFit LT software (version 3.0, Verity Software House, Topsham, ME). We used calibration standards (Molecular Probes, Carlsbad, CA) and a DNA QC particle kit (Becton Dickinson) to verify instrument performance.
Statistics-The data were analyzed using the GraphPad Prism 5.0 software (GraphPad Software, Inc., La Jolla, CA). Student's t test was applied to determine statistical significance between the treatments. Unless indicated otherwise, the level of statistical significance was set at p Ͻ 0.05.

RESULTS
hMLH1 Complementation Sensitizes HCT 116 Cells to Selenium Compounds and Potentiates the Selenium-induced DNA Oxidation and Breaks-Selenium-induced cell cycle arrest and apoptosis in colon cancer cells has been well studied (28 -30); however, a direct role of hMLH1 in seleniummediated DNA damage response has not been studied under an isogenic background. To determine whether hMLH1 plays a role in the response of the cancerous cells to selenium compounds, we first assessed sensitivity of HCT 116 and HCT 116ϩhMLH1 cells to Na 2 SeO 3 (Fig. 1A), MSeA (Fig.  1B), and MSeC (Fig. 1C) by colony formation assays. We chose the selenium compounds because the organic MSeA and MSeC are known to be very efficacious in suppressing tumors in animal models (31), yet Na 2 SeO 3 is as effective as the organic selenium compounds in activating early barriers of tumorigenesis in cell models (22). Here, we found that HCT 116ϩhMLH1 cells were significantly (p Ͻ 0.05) more sensitive than HCT 116 cells to the three selenium compounds in a dose-dependent manner. This observation was confirmed by 3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide assays (supplemental Fig. S1) and sub G 0 /G 1 cell profiles containing fragmented DNA (HCT 116ϩhMLH1 cells, 7-10%; HCT 116 cells, trace amount) after selenium exposure (supplemental Fig. S2). Consistent with the notion that selenium compounds can induce ROS formation (22,32), addition of the antioxidant NAC and Tempo dramatically suppressed the selenium-induced cell death in HCT 116ϩhMLH1 (Fig. 1, D-F) and HCT 116 cells (supplemental Fig. S3) as evidenced by colony formation assays. The expression of hMLH1 in the HCT 116ϩhMLH1 cells was comparable to that in the hMLH1-proficient Caco-2 colorectal cancer cells but was substantially lower than the HeLa cervical cancer cells and higher than the noncancerous CCD841 colorectal cells (supplemental Fig. S4). This suggests that the hMLH1 expression in the HCT 116ϩhMLH1 cells is not hyperexpressed and is at a physiological level.
We next determined the nature of DNA damage in the selenium-treated colorectal cancer cells. The mutagenic 8-oxoG oxidative DNA lesions can lead to GC 3 TA mutations and DNA strand breaks if left unrepaired (33). H2AX is a phosphorylation substrate of ATM, resulting in the expression of ␥H2AX, a marker of DNA breaks. Analyses of immunofluorescence results indicated that 8-oxoG (Fig. 2, A and B) and ␥H2AX (Fig. 2, C and D) in the nuclei were induced after treatment with Na 2 SeO 3 (2 M), and the extent was much greater (p Ͻ 0.05) in HCT 116ϩhMLH1 than in HCT 116 cells. Pretreating HCT 116ϩhMLH1 cells with the antioxidant NAC or Tempo significantly (p Ͻ 0.05) reduced the Na 2 SeO 3 -induced 8-oxoG and ␥H2AX formation (supplemental Fig. S5). Altogether, hMLH1 sensitizes HCT 116 cells to selenium compounds in a manner dependent on ROS and promotes oxidative 8-oxoG damage and DNA breaks.
hMLH1 Complementation Facilitates G 2 /M arrest in Selenium-treated HCT 116 Cells-Previous results indicate a selenomethionine dose-dependent cell cycle arrest in HCT 116 cells (29,30). Here, we determined a role of hMLH1 in selenium-induced DNA damage checkpoint response in colorectal cancer cells. Flow cytometric analyses indicated a time-dependent G 2 /M arrest 0 -12 h after Na 2 SeO 3 exposure (2 M) (Fig. 3C) in HCT 116ϩhMLH1, but not in HCT 116 cells. The G 2 /M arrest in HCT 116ϩhMLH1 cells subsided at 24 and 48 h. Consequently, G 0 /G 1 (Fig. 3A) and S (Fig. 3B) populations were less abundant in HCT 116ϩhMLH1 than in HCT 116 cells at 6 and 12 h. Similar patterns of cell cycle profiles exist 0 -48 h after cellular exposure with MSeA (2 M) or Na 2 SeO 3 (5 M) (supplemental Fig. S6). The above studies implicate hMLH1 in G 2 /M cell cycle arrest in selenium-treated HCT 116 cells.
Selenium-induced G 2 /M Arrest in HCT 116ϩhMLH1 Cells Depends on ATM Kinase Activity and ROS-We have shown previously that selenium-induced DNA damage response involves ROS and ATM (22). ATM activation initiates a cascade of DNA damage responses, including G 2 /M arrest, to various types of DNA damage (17)(18)(19). Therefore, we treated HCT 116ϩhMLH1 cells sequentially with the ATM kinase inhibitor KU55933, the antioxidant NAC, or the superoxide dismutase mimic Tempo and Na 2 SeO 3 (2 M). The results showed that pretreatment of the cells with KU55933 (Fig. 4A), NAC, or Tempo (Fig. 4B) attenuated the selenium-induced G 2 /M arrest. KU55933, NAC, or Tempo alone did not impact on cell cycle profiles in HCT 116ϩhMLH1 cells.

Selenium-induced ATM Pathway Activation Requires hMLH1
and Occurs Mainly in G 2 /M Phase-The above results suggest that hMLH1 may mediate selenium-induced ATM pathway activation in the G 2 /M phase. To test this hypothesis, we determined the expression of G 2 /M phase pATM Ser-1981, a general marker of ATM pathway activation (34), and pDNA-PK cs Thr-2647 (phospho-DNA-PKcs on Thr-2647), a phosphorylation event downstream of ATM kinase activation (35), in HCT 116   OCTOBER 22, 2010 • VOLUME 285 • NUMBER 43 and HCT 116ϩhMLH1 cells treated with Na 2 SeO 3 (2 M) with or without pretreatment of the microtubule blocker Noc (300 ng/ml) for 16 h. This condition traps about 90% cells in G 2 /M phase (36,37). Immunofluorescence analysis showed that pATM Ser-1981 (Fig. 5, A and B) and pDNA-PK cs Thr-2647 (Fig. 5, C and D) levels increased marginally in HCT 116 cells after selenium exposure. In stark contrast, both the phosphorylation events are substantially induced in HCT 116ϩhMLH1 6 and 12 h after the Na 2 SeO 3 exposure. Furthermore, the Na 2 SeO 3 -induced expression of pATM Ser-1981 and pDNA-PK cs Thr-2647 were significantly enhanced by Noc treatment in HCT 116ϩhMLH1 but not in HCT 116 cells. Noc treatment alone did not induce expression of pATM Ser-1981 or pDNA-PK cs Thr-2647. Pretreatment of HCT 116ϩhMLH1 cells with NAC, Tempo, or KU55933 significantly (p Ͻ 0.05) attenuated Na 2 SeO 3 -induced pATM Ser-1981 formation (supplemental The Association between hMLH1 and hPMS2 Is Stimulated by Selenium Treatment and Depends on ATM Kinase Activity and ROS-At the early stage of MMR, hMLH1 and hPMS2 form a heterodimer to stabilize mismatched DNA and facilitate DNA damage response (13). Therefore, we performed co-immunoprecipitation experiment to determine whether the association between hMLH1 and hPMS2 was altered after selenium treatment. We found that treatment of HCT 116ϩhMLH1 cells with Na 2 SeO 3 (Fig. 6A) and MSeA (supplemental Fig. S8) increases the amount of hMLH1 in the anti-hPMS2 immunoprecipitates at 24 h in a dose-dependent manner. Reciprocal co-immunoprecipitation was not achievable in our hands  using available anti-hMLH1 antibodies. Incubation of the immunoprecipitates with ethidium bromide did not affect the protein associations (supplemental Fig. S9), suggesting that the protein-protein interaction was not due to DNA bridging. Next, we determined whether ROS and ATM kinase were involved in the increased hMLH1-hPMS2 interaction. Pretreatment with NAC (Fig. 6B), Tempo (Fig. 6C), or KU55933 (Fig. 6D) prevented the hMLH1-hPMS2 association in HCT 116ϩhMLH1 cells with or without selenium treatment. Similar to a previous report (25), the selenium compounds, antioxidants, and KU55933 treatment had no effect on protein levels of hMLH1 or hPMS2 (Fig. 6, A-D and supplemental Fig. S8). Taken together, selenium treatment stimulates hMLH1-hPMS2 association in a manner dependent on ATM kinase and ROS.

DISCUSSION
In this study, we have identified hMLH1 as a selenium target for ATM-dependent G 2 /M arrest in colorectal cancer cells and provided insight into the resistance of the cancer cells to selenium treatment. Selenium may mitigate colorectal tumorigenesis at various stages by different mechanisms. Our recent study provides evidence that selenium compounds at low doses can activate an ATM-and ROS-dependent senescence response, an early barrier of tumorigenesis, in the noncancerous CCD841 but not the cancerous HCT 116 colorectal cells (22). The current study describes an essential role of hMLH1 in the ATM-dependent DNA damage response in the G 2 /M phase of HCT 116 cells exposed to relatively high doses of selenium. We show that hMLH1 complementation sensitizes HCT 116 cells to selenium-induced oxidative and broken DNA damage and stimulates the ATM-dependent G 2 /M checkpoint response in HCT 116 cells. After selenium exposure, the ATMdependent DNA damage response promotes the hMLH1-hPMS2 association that is critical to initiate MMR. Without the hMLH1-depenedent ATM pathway activation, the cancerous HCT 116 cells fail to effectively transmit the selenium-induced oxidative DNA damage signal to checkpoint and cell death pathways. Consequently, the colorectal cancer cell may proliferate with unrepaired DNA damage and exacerbate the intrinsic genome instability.
How does hMLH1 complementation sensitize HCT 116 cells to selenium-induced cell killing effect? Although the primary function of MMR is to correct errors arising during DNA replication, MMR proteins are known to strongly enhance cellular and genetic toxicity of several chemotherapeutic agents (38,39). Metabolism of selenium compounds at chemopreventive or therapeutic doses are known to induce ROS formation and the subsequent oxidative DNA base modifications and breaks in cancerous cells, including HCT 116 cells (22, 32, 40 -45). MMR may mediate selenium-induced DNA breaks by two mechanisms. Firstly, the model of futile MMR cycle depicts that the generation of single-stranded breaks and gaps after the first S-phase can be converted into DNA DSBs by replication fork collapse when the unrepaired DNA lesions enter the next cycle of S-phase (38,39). Secondly, recognition and processing of DNA damage by MMR proteins may directly induce DNA DSBs (46,47). Because the selenium-induced ␥H2AX formation appears as fast as 8-oxoG formation, our study favors the latter scenario in which hMLH1 plays a signaling but not repair role in recognizing and processing the selenium-induced oxidative DNA damage in G 2 /M phase. The hMLH1-hPMS2 complex may be recruited to the region containing 8-oxoG, and its intrinsic nuclease activity may process the oxidatively modified DNA into breaks (48). Similarly, in the cellular response to the carcinogenic chromium, the MMR signaling complexes MSH2-MSH3 and MSH2-MSH6 contribute to a rapid formation of chromium-induced DNA DSBs (47). Consequently, the selenium-induced HCT 116ϩhMLH1 cells accumulate the unrepaired and highly toxic DNA breaks, arrest in G 2 /M phase, and enter the cell death pathway.
The ATM kinase is crucial in the response to DNA breaks and then transmits the signal to downstream substrates for S and G 2 /M arrest and/or cell death. An hMLH1-dependent ATM pathway activation has been demonstrated (20,21), and cells deficient in MMR have defective G 2 /M arrest after DNA DSBs (16). As discussed above, the hMLH1-hPMS2 complex may rapidly process oxidative DNA damage into DNA breaks, which in turn activates the ATM kinase. Similar to chromiuminduced DNA DSBs (46), the selenium-induced lesions are found primarily in G 2 /M phase. The nature of the DNA DSB induction by chromium and selenium is not totally understood, but this may involve cell cycle-specific assembly of DNA DSBsensing protein complex such as the endonuclease-containing Mre11/Rad50/Nbs1 complex that can be recruited to damaged site in the S and G 2 phase (49). In particular, the Mre11/Rad50/ Nbs1 complex interacts with hMLH1 in G 2 phase after the formation of O(6)-methylguanine induced by the chemotherapeutic agent temozolomide (50). Whatever the attribution, the hMLH1-dependent DNA DSB induction by selenium may activate ATM kinase in G 2 phase, which in turn signals apoptosis pathway through key players such as p53, a prominent phosphorylation substrate of ATM (51). Interestingly, immunohistochemistry analysis of colorectal cancer samples reveals that expression of ATM is associated with cancer survival (52).
What does the increased association between hMLH1 and hPMS2 in HCT116ϩhMLH1 cells after selenium exposure implicate? To the best of our knowledge, this is the first report showing an increased association between hMLH1 and hPMS2 after DNA damage. hPMS2 is not expressed in the hMLH1deficient HCT116 cells, suggesting the requirement of hMLH1-hPMS2 complex for functional MMR in colorectal cancer cells. Interestingly, ATM has been shown to interact with and possibly phosphorylates hMLH1 (53). Consistent with this notion, we found the increased hMLH1-hPMS2 association by selenium treatment is dependent on ATM kinase activity. ATM activation after DNA damage may stimulate hMLH1-hPMS2 association as observed in the case of an increased association between hPMS2 and p73 protein in the apoptotic response to cisplatin-induced DNA damage (25). Moreover, hMLH1 and hPMS2 both exhibit nuclear import and export signal domains, but hMLH1-hPMS2 dimerization masks the export peptides located at the C terminus (54). It is conceivable that ATM kinase activity facilitates the association between hMLH1 and hPMS2 and also confines the complex in the nucleus for efficiently sensing and processing the selenium-induced oxidative DNA damage.
In animal tumor models, MSeA and MSeC are more efficacious than Na 2 SeO 3 in suppressing chemical-induced tumori-genesis (31). Nonetheless, LD 50 of HCT 116ϩhMLH1 cells by MSeC is 100-fold higher than that by MSeA or Na 2 SeO 3 . Cultured cells lack a lyase and other enzymes necessary for conversion of MSeC to more effective downstream selenium metabolites (55,56). Therefore, future studies employing selenium compounds in cultured cells for chemoprevention perspectives should avoid using MSeC.
Altogether, we propose that the hMLH1-hPMS2 complex senses and processes selenium-induced oxidative DNA damage and transmits the signal to ATM kinase, leading to the activation of G 2 /M checkpoint and death pathways. Nonetheless, other kinases such as DNA-PKcs also can be involved in the hMLH1-dependent sensitivity to selenium compounds. Because pDNA-PKcs Thr-2647 induction by DNA DSBs requires ATM (35), hMLH1 may mediate the seleniuminduced formation of pDNA-PKcs on Thr-2647 via ATM. In HCT116cells,lossofafunctionalMMRcompromisestheATMdependent G 2 /M checkpoint response and contributes to resistance to selenium compounds and tolerance to aberrant DNA adducts. Consequently, cells may avoid apoptosis and enter mitosis with persistent DNA damage, exacerbating genome instability in the cancer cell. Although the majority of colorectal cancer progressively loses MMR function, activation of the hMLH1-dependent G 2 /M checkpoint response by selenium is a promising therapeutic target in precancerous or hMLH1-expressing colorectal cancer cells. Along with our recent publication (22), we propose that selenium compounds can differentially activate DNA damage response at various stages of colorectal tumorigenesis. In the initiation stage, DNA damage induced by selenium compounds at sublethal doses activates a senescence response (22), thus stifling cancer progression at a very early stage. During the promotion and progression stages of tumorigenesis, selenium compounds at lethal doses halt cells in the G 2 /M phase and promote cell death in a hMLH1-depdenent manner. The identification of the hMLH1dependent activation of ATM DNA damage response provides mechanistic insight into selenium chemoprevention on colorectal cancer and suggests a promising molecular pathway for selectively killing of cancer cells by selenium.