Many anticancer agents kill cells as a consequence of damaging their DNA. While this approach targets rapidly proliferating cells, it is only somewhat selective for tumors and can lead to undesirable toxicity to the patient. Differences exist between normal and tumor cells with respect to the cellular response to DNA damage and it may be possible to exploit these differences to selectively target tumor cells. One such difference is in cell cycle checkpoint regulation. In response to DNA damage, cell cycle checkpoints are activated causing the arrest of cell cycle progression [1]. This allows the cell time to repair the damage, or if too much damage has occurred, undergo cell death. Defects in checkpoints lead to mutations and cancer, and this genomic instability is now considered a characteristic of tumors [2]. As the checkpoints function to protect the cells from damage, it is hoped that their selective defect in tumor cells may result in sensitivity to novel therapeutic strategies.

The S-phase checkpoint is activated in response to DNA damage that occurs during replication. For example, topoisomerase I inhibitors such as SN38 produce a single-strand break in DNA which is converted to a double-strand break upon collision with the replication fork [3]. The MRN complex, consisting of Mre11, Rad50 and Nbs1, is recruited to the double-strand break and initiates the S-phase checkpoint by recruiting ataxia-telangiectasia-mutated (ATM) to the damaged site [4]. Processing at the break by the Mre11 nuclease generates regions of single-stranded DNA that recruit the ATM- and Rad3-related kinase (ATR) [5]. Activated ATR and ATM phosphorylate the checkpoint kinases Chk1 and Chk2, respectively, which in turn phosphorylate CDC25A targeting it for degradation and effectively arresting DNA synthesis through inactivation of the Cdk2/Cyclin A complex [6, 7].

Mutations in the MRN complex have been identified in lymphoid, breast and colorectal tumors [8, 9]. A defect in the MRN complex has also been identified in the colorectal tumor cell line, HCT116 [9]. Such mutations result in radioresistant DNA synthesis (RDS) and hypersensitivity to ionizing radiation [10–12]. In addition, MRN mutations have been reported to increase sensitivity towards anticancer drugs, such as topoisomerase I inhibitors and cisplatin [13, 14].

Studies in our laboratory have confirmed that HCT116 cells have very low levels of Mre11 and also that they arrest poorly in S phase in response to SN38. By transfecting HCT116 cells with a vector expressing Mre11, we show here that we can reconstitute the MRN complex. Our data indicate that Mre11 is necessary for the stability of Rad50 and Nbs1 and rescues the S-phase arrest in HCT116.

MRN defects if present in some tumors could render these cells sensitive to DNA-damaging agents and thereby provide a tumor selective therapy. The primary goal of these studies was to find compounds that preferentially target cells with defects in the proteins of the MRN complex. We have examined the frequency of MRN defects across the NCI60 cell line panel which is derived from many different cancer types (albeit only 59 are presently available) [15]. These cell lines have been characterized pharmacologically by exposure to more than 100,000 compounds (http://dtp.nci.nih.gov). We then correlated the MRN protein levels to sensitivity to the compounds in the NCI database using the COMPARE program [16, 17]. Since there are many other proteins involved in the S-phase checkpoint we also assessed the cell lines for their ability to arrest in S phase when incubated with SN38, and then searched the database for additional correlations.

The overall goal of this study was to identify molecular defects in cell cycle checkpoint regulation using the NCI60 panel of cell lines, and then to identify compounds that might target cells with these defects. Eventually, such compounds might provide leads for the development of drugs for personalized cancer therapy. We began by screening the cell panel for potential defects in MRN proteins and the S-phase checkpoint. We uncovered a number of interesting relationships with respect to MRN proteins and their mRNA that impact the level of this complex. In addition, using the COMPARE program, we have identified several compounds of interest which warrant further study.

Our results implicate protein stability in the regulation of MRN protein levels. Transfection with Mre11 not only increased Mre11 levels but also Rad50 and Nbs1 levels. In comparing the protein levels in each cell line across the NCI60 panel, we also found a strong correlation between the levels of these three proteins. The overall level of the complex appeared to be primarily regulated by transcription of Mre11 as there was much lower correlation between mRNA and protein levels for Rad50 and Nbs1. These results suggest that the protein levels of Rad50 and Nbs1 are primarily regulated by stability of the complex while transcription of Mre11 contributes more to the overall level of the complex.

While Mre11 expression may be more highly regulated by transcription, among the cell lines with low levels of Mre11 protein, only the IGROV1 cells had a very low level of mRNA. HCT116 had a relatively high level of Mre11 transcript but low protein, and this has been shown to be due to a splicing defect in its mRNA [9]. This defect occurs as a frame shift in a poly(dA) track in an intron which is consistent with these cells being defective for mismatch repair MLH1 [9, 24]. Five cell lines in the NCI60 panel are defective for MLH1 [25], one of which is IGROV1, but the other three (SKOV3, KM12 and CCRF-CEM) still express relatively normal levels of Mre11. None of the NCI cell lines are defective in the mismatch gene MSH2. Hence mismatch repair defects do not appear to explain the low levels of Mre11 protein in these other cell lines. Other mutations which might result in protein degradation could be a possible cause for the low protein levels in the other cells but further study is necessary. In addition, there were several cell lines with unbalanced levels of MRN, for example, unexpected high or low levels of Rad50 compared to Mre11. One possible explanation for this observation is that the proteins contain a mutation which confers greater protein stability when they are not in the MRN complex.

This screen identified a small number of compounds which correlated with either Mre11, Rad50, or Nbs1. However, there were three compounds identified which correlated with both Mre11 and Rad50. The activity of one of these, NSC 734406, was significant at the concentrations tested and correlates with low levels of Mre11 and Rad50 (Figure 4). Validation and further study with this compound is necessary to assess whether it does preferentially kill cells with low MRN protein but this was not possible as the compound is not available from the NCI repository. One of the potential limitations of this COMPARE analysis is that protein levels may not necessarily reflect MRN function, particularly in those cases where MRN protein expression is unbalanced. Another approach would be to assay the cells for MRN activity using a functional readout, such as Chk1 phosphorylation in response to ionizing radiation, for which the MRN complex is required [26]. Alternately the recruitment of ATM to damaged DNA could be assayed as this also depends on the MRN complex.

In searching the NCI60 panel for cell lines with defects in the S-phase checkpoint we identified three cell lines, in addition to HCT116, which failed to arrest in S-phase in response to SN38. HCT116, TK-10 and IGROV1 all have low levels of the MRN which may explain their S phase defect (see summary in Figure 10). In contrast, HCC2998 cells fail to arrest in S phase yet have very high levels of Rad50 and Mre11 proteins. It remains to be determined whether the MRN complex in this cell line is functional. If functional, this would suggest that while the MRN complex is necessary for S-phase arrest, it is not sufficient. There were also several cell lines with low levels of the MRN proteins that were proficient for S-phase arrest. A possible explanation is that low levels of MRN maybe sufficient for S-phase arrest but that a mutation in Mre11 as occurs in HCT116 may be necessary to elicit the checkpoint the defect. For example, it has been suggested that the mutant form of Mre11 in HCT116 acts in a dominant negative manner and this may be far more detrimental than simply having low levels of Mre11 [24].

We have previously reported an alternate means to prevent S phase arrest [27]. The expression of shRNA against Chk1 in MDA-MB-231 cells overcame the S phase arrest induced by SN38, although such cells still arrested in G2. We observed a similar response when cells were co-incubated with SN38 and a small molecule inhibitor of Chk1. There is a close link between the MRN complex and Chk1 activation. Specifically, MRN is recruited to double-strand breaks where the exonuclease activity of Mre11 creates single-stranded DNA that, through a series of events, recruits ATR and activates Chk1. However, it has been shown that HCT116 cells are fully competent to activate Chk1 when damaged with topoisomerase I inhibitors [21] and we have confirmed this in our own laboratory (data not shown). This is likely explained by the topoisomerase I inhibitors stalling replication fork progression thereby activating ATR and Chk1 without the need to implicate ATM and Mre11. This raises the conundrum as to why SN38 is fully capable of activating Chk1 in HCT116 cells yet the cells fail to S phase arrest unless they are complemented with Mre11. The results suggest that activation of Chk1 may be necessary but not sufficient to induce S phase arrest and that the MRN complex plays some additional role.

In searching for correlations between S-phase checkpoint defects and compound activity, we identified many drugs which correlated positively, meaning these compounds have activity in cell lines with defects. Six compounds out of the top 30 correlations identified mapped to the same cluster indicating that they share the same mechanism of growth inhibition. Four of these are benzothiazoles, a class of compounds that has been shown to be metabolized by sensitive cells and then form DNA adducts [22, 23]. Whether this is also true of other compounds in this cluster has not been determined. However, we note that KM12 cells have previously been shown to metabolize benzothiazoles, yet they are insensitive to these compounds [28]. Similarly we note that TK10 cells are as sensitive as MCF7, yet have markedly lower level of metabolism [22]. The lack of sensitivity of HCT116 to benzothiazoles may be due to their inability to metabolize the compound, however HCT116 cells (and many others) remained insensitive to benzothiazoles when co-cultured with MCF7 cells. Accordingly, while metabolism may contribute to the sensitivity to these compounds in some cells, there may also be a major contribution from the S-phase checkpoint defect.

The fact that the S-phase checkpoint search was a functional assay provides greater confidence in the candidates identified than with the MRN protein screen. However, because the S-phase defect can be caused by multiple reasons (e.g., low or defective ATM, ATR, Mre11 or Chk1), we may have missed many potential drug candidates. Some compounds may correlate with only some of the checkpoint defects responsible for the lack of SN38-induced S-phase arrest. The fact that each defect may be rare in the NCI60 panel makes it a challenge to use this approach to identify novel drug candidates. For example, a compound that has the same activity in 60 cell lines would exhibit a very high correlation with a dataset in which 56 cell lines arrested and 4 did not, the correlation coefficient being driven by the 56 cell lines rather than the 4 interesting queries. Accordingly, each compound discovered in the screen had to be individually analyzed and validated. Many compounds elicited no obvious correlation with the query set. In some cases (e.g., NSC358873), three of the four cell lines that failed to arrest in S phase had not even been screened against this compound. These limitations are unfortunate because we had hoped to use this approach to identify compounds that killed only a few cell lines, and might therefore be candidates for selective therapy for tumors with that particular phenotype (i.e., individualized therapy). Given the limitations of this screen, the fact that most of the cells defective for S phase arrest were indeed sensitive to compounds in the benzothiazole cluster gives strong support for further investigation into this class of drugs.

Genomic instability is one of the characteristics of cancer and is frequently attributable to defects in DNA damage response genes involved in either DNA repair or checkpoint regulation. This provides a therapeutic opportunity as these defective cells may be hypersensitive to various drugs. We have identified cells with potential defects in the MRN complex and in S phase arrest, and a series of compounds that may preferentially target S phase arrest-defective cells. The in silico screen used has limitations when seeking compounds that are cytotoxic to only a few cell lines. But considering this limitation, the identification of benzothiazoles as compounds that preferentially target cells with a defect in S phase arrest warrants further investigation. Compounds such as these that kill only a few cell lines likely represent the future of individualized cancer therapeutics.