Competitive and Noncompetitive Inhibition of the DNA-dependent Protein Kinase¹

DNA-PK³ is a serine/threonine protein kinase that is relatively abundant in human cell nuclei (1, 2). A notable feature of this enzyme is that it is activated on binding to DNA lesions, such as double-strand breaks, nicks, and DNA single single-to-double-strand transitions (3). DNA-PK is composed of a large (M_r ∼470,000) catalytic subunit, DNA-PKcs, and a DNA targeting component, termed Ku (4, 5, 6). Ku, first identified as a human autoimmune antigen, is a heterodimer composed of polypeptides of M_r ∼70,000 and ∼80,000 (Ku70 and Ku80, respectively). Both components of DNA-PK function in the repair of DNA damage induced by IR and, hence, show a reliance on each other for generating a functional holoenzyme in vivo (7, 8, 9). Cells defective in Ku components (such as xrs-6 and lines derived from Ku70^−/− and Ku80^−/− mice) are not only sensitive to DNA double-strand break-inducing agents but are also defective in the site-specific nonhomologous end-joining process of V(D)J recombination (7, 8, 9). The severe combined immunodeficient mouse possesses a truncated DNA-PKcs polypeptide that appears to lack kinase activity, and this also results in defective V(D)J recombination and an inability to effectively repair IR-induced DNA double-strand breaks (10, 11, 12). It has been shown recently that DNA-PKcs can bind to and be activated by DNA in a Ku-independent manner in vitro (13, 14). However, because Ku-deficient cells show similar phenotypes to DNA-PKcs-deficient cells, it is likely that Ku is required for DNA-PK function in physiological situations.

Cloning of the DNA-PKcs cDNA revealed that its open reading frame directs the synthesis of a protein of ∼4100 amino acid residues (6). Other than a leucine zipper motif found between residues 1503 and 1538, which appears to be involved in interacting with the nuclear matrix protein C1D (15), the only clearly identifiable functional motif in the DNA-PKcs polypeptide is the kinase domain comprising the ∼400 COOH-terminal amino acid residues. This region shows very little homology to classical serine/threonine kinases, but it shows much greater sequence similarity with the catalytic domains of proteins involved in the phosphorylation of PI, notably PI 3-kinases and PI 4-kinases (16). Other polypeptides shown to contain this PI 3-kinase-like domain include several proteins implicated in DNA repair, genome surveillance, and cell cycle checkpoint controls: mammalian ATM, ATR, and FRAP, together with homologues of these in systems such as Saccharomyces cerevisiae and Schizosaccharomyces pombe (17). To date, members of this latter group have not been shown unequivocally to possess an intrinsic ability to phosphorylate phosphatidlyinositols and are, therefore, most probably protein serine/threonine kinases (18).

Wortmannin is a seterol-like fungal metabolite that has been shown previously to be a potent and selective inhibitor of mammalian PI 3-kinase (19, 20). On the basis of the similarity between the kinase domains of DNA-PKcs and PI 3-kinases, we and others have established that wortmannin is able to inhibit DNA-PK function in vitro (6, 21, 22, 23). However, the IC₅₀ (concentration at which 50% of activity is lost) for wortmannin toward DNA-PK is, at 250 nm, ∼2 orders of magnitude higher than that for PI 3-kinase. Nevertheless, this observation has led to the demonstration by other groups that cells grown in the presence of wortmannin are sensitized to the killing effects of IR (24, 25, 26, 27). Therefore, in this context, wortmannin acts as a radiosensitizing agent. Although the above effects suggest that the radiosensitization is via DNA-PK inhibition, the possibility exists that wortmannin also impairs the function of other members of the PI 3-kinase family, such as ATM or ATR (17, 18, 23, 28). More recently, it has been shown that another PI 3-kinase inhibitor, LY294002, is also able to sensitize cells to the effects of IR (26).

Because of the potential use of wortmannin, LY294002, and their derivatives as radiosensitizers, we have sought to provide a detailed analysis of their inhibition of DNA-PK. Here, we provide a kinetic analysis of the PI 3-kinase inhibitors wortmannin, quercetin, quercitrin, rutin, and LY294002 toward DNA-PK. Furthermore, we have used wortmannin as a probe for DNA-PK function in vitro. Our data not only provide a detailed analysis of DNA-PK inhibitors in vitro but also serve to broaden the known effects of wortmannin and quercetin derivatives. We are also able to show here that wortmannin can bind directly to DNA-PKcs in vivo. Our data provide a starting point for the development of more specific inhibitors of PI 3-kinase family members and potentially novel radiosensitizers.

The human lymphoblastoid cell line HL60 was obtained from American Type Culture Collection (ATCC CCL-24; Manassas, VA) and maintained in RPMI 1640 supplemented with 10% FCS, glutamine, and antibiotics. The human glioma cell lines MO59J and MO59K (29), kindly provided by J. Allalunis-Turner (Cross Cancer Institute, Edmonton, Alberta, Canada) were grown in DMEM: Ham’s F-12 medium, supplemented with 10% FCS, glutamine, and antibiotics.

Wortmannin, quercetin, and LY294002 were obtained from Alexis Corporation (UK) Ltd. (Nottingham, United Kingdom). Rutin and quercitrin were purchased from Sigma (Poole, United Kingdom).

DNA-PK assays were performed as described previously using a p53-based peptide substrate (30, 31). For inhibition experiments, wortmannin was added to the reaction mixture at the same time as ATP. For the other inhibitors, these were preincubated with DNA-PK for 5 min prior to ATP addition. All chemical inhibitors were dissolved in 10% DMSO, resulting in a final DMSO concentration of 1% in the assay mixture. DMSO (1%) was found not to adversely affect the DNA-PK assay.

DNA-PK holoenzyme and DNA-PKcs were purified from HeLa nuclear extract (Computer Cell Culture Centre, Mons, Belgium) essentially as described previously (32).

Nuclear extract (10 μg), purified DNA-PK (600 ng), or DNA-PKcs (100 ng) was labeled with wortmannin (10 μm) at room temperature for 20 min in buffer containing 5 mm HEPES (pH 7.6), 40 mm KCl, 4% glycerol, 0.2 mm MgCl₂, and 1 mm DTT. Wortmannin-labeled proteins were identified by standard SDS-PAGE (7% polyacrylamide gels) and Western blotting analysis with the use of antiwortmannin antiserum kindly provided by M. Wymann (Fribourg, Switzerland; Ref. 33). In vivo labeling of cells was carried out by adding wortmannin to a final concentration of 10 or 100 μm to the growth medium for 90 min. Nuclear extract was prepared from harvested cells by the method of Dignam et al. (34). Wortmannin-labeled proteins were identified as outlined above. For limited proteolysis of DNA-PKcs, 1 or 3.3 μl of recombinant caspase-3 (Pharmingen, San Diego, CA) was added to 600 ng of DNA-PK labeled with 10 μm wortmannin (as above) and incubated at 37°C for 30 min prior to Western blot analysis using anti-DNA-PKcs antiserum (AbFLA; Ref. 35) or antiwortmannin antiserum (33).

Because PI 3-kinase activity is inhibited potently by wortmannin and because DNA-PKcs shares homology across its kinase domain with PI 3-kinases, we tested previously whether wortmannin was able to inhibit DNA-PK activity (6). This earlier study showed that wortmannin is able to inhibit DNA-PK activity, with an IC₅₀ of ∼250 nm. However, this study did not ascertain the mode of inhibition by wortmannin or provide an accurate K_i. Thus, we assayed DNA-PK activity at a variety of different ATP concentrations in the presence of increasing concentrations of wortmannin. A Lineweaver-Burk plot of this analysis revealed that wortmannin behaves as a noncompetitive inhibitor for the ATP binding site of DNA-PK (Fig. 1,A). From a Dixon plot of this analysis (reciprocal of the velocity of the reaction plotted against wortmannin concentration), a K_i of 120 nm was obtained (Fig. 1,B). Because the primary plot of these data could not fully discriminate whether the nature of wortmannin inhibition was noncompetitive or mixed noncompetitive/uncompetitive inhibition, Eadie-Scatchard analysis of the initial data set was performed (Fig. 1 C; Ref. 36). By plotting the velocity of reaction against velocity over substrate concentration, we show, from the slope of the graphs that are constructed, that the mode of wortmannin inhibition is essentially noncompetitive in nature (36).

Like wortmannin, LY294002 has been found previously to be an inhibitor of PI 3-kinase (37). With this in mind, we sought to determine whether this compound is able to inhibit DNA-PK activity and, if so, to determine its mode of action. To study the possible effects of LY294002, we carried out DNA-PK assays in a variety of different ATP concentrations in the presence of increasing amounts of LY294002. The Lineweaver-Burk plot constructed from this analysis revealed that LY294002 acts as a competitive inhibitor for the ATP binding site of DNA-PK (Fig. 2,A). By plotting the slopes from the primary graph (Fig. 2,A) against the concentration of inhibitor, we calculated a K_i of 6.0 μm (Fig. 2 B). The competitive nature of LY294002 against ATP for DNA-PK is, thus, in keeping with its mechanism of inhibition for PI 3-kinase (37).

LY294002 was initially identified through a screen of compounds derived from quercetin to identify more specific inhibitors of PI 3-kinase (37). The bioflavanoid quercetin has also been shown previously to inhibit a variety of kinases, including PI 3-kinase and PI 4-kinase (38, 39). Hence, we investigated whether quercetin and the related flavanoids quercitrin and rutin are able to inhibit DNA-PK. Lineweaver-Burk analysis revealed that all three compounds inhibit DNA-PK activity and do so by competing with ATP in a competitive manner, similar to LY294002 (data not shown). Secondary plots of the slopes from the primary graph against the inhibitor concentrations used (data not shown and Table 1) revealed that rutin has a K_i of 26 μm, quercetin has a K_i of 110 μm, and quercitrin has a K_i of 208 μm for DNA-PK.

The development of antisera to wortmannin has allowed the performance of elegant studies to identify exactly where this compound reacts in the active site of the M_r 110,000 subunit of mammalian PI 3-kinase (33). Thus, after formation of wortmannin adducts with PI 3-kinase followed by limited proteolysis, analysis of the resulting peptides with the antiwortmannin antibody and by mass spectrometry revealed that wortmannin targets a highly conserved lysine residue in the active site of PI 3-kinase. Sequence analysis reveals the conservation of this amino acid residue in DNA-PKcs. Therefore, we wanted to see whether covalent binding to DNA-PK by wortmannin could be observed using the antiwortmannin antibody. Because effective detection of wortmannin-labeled PI 3-kinase required levels of wortmannin ∼100-fold above its IC₅₀ (33), we used a range of wortmannin concentrations of up to 10 μm in our initial studies. Notably, and as shown in Fig. 3,A, over a wide range of concentrations, wortmannin does, indeed, form covalent adducts with DNA-PKcs that can be detected by Western blot analysis using the antiwortmannin antibody. Furthermore, as indications of the specificity of this reaction, the two Ku subunits are not targeted by wortmannin, and the formation of wortmannin adducts to DNA-PKcs is competitively inhibited by ATP (Fig. 3 B).

It has recently been shown that DNA-PKcs is cleaved during apoptosis by caspase-3 or a related enzyme (35, 40, 41). Thus, caspase-3 has been shown to proteolyse DNA-PKcs into a large M_r 240,000 NH₂-terminal region (Δ240; Fig. 3,C) and COOH-terminal kinase domain portions of M_r 120,000 and 150,000 (Δ120 and Δ150; Fig. 3,C). As an approach to help define the site of wortmannin adduct formation on DNA-PKcs, we, therefore, used recombinant caspase-3 to perform limited proteolysis on DNA-PKcs that had been pretreated with wortmannin. As revealed by probing with AbFLA, a polyclonal antiserum raised against the whole DNA-PKcs polypeptide, caspase-3 treatment cleaves DNA-PKcs into the fragments described previously (Fig. 3 C, left). More significantly, probing the same material with the antiwortmannin rabbit polyclonal antiserum reveals that only the COOH-terminal regions of DNA-PKcs that contain the kinase domain are labeled by wortmannin. We, therefore, conclude that wortmannin forms covalent adducts specifically to the kinase region of DNA-PKcs but not significantly to any other region of the protein.

Next, we used wortmannin and the specific antiserum to this compound to address a variety of questions regarding the status of the DNA-PKcs active site under various circumstances. Previous work has established that DNA-PKcs has an inherent capacity to bind to DNA termini and is activated by the simultaneous presence of the non-sequence-specific DNA-binding protein Ku and DNA ends. To address whether Ku activates DNA-PKcs via opening of the ATP-binding site of the kinase, we added wortmannin to a DNA-PKcs preparation that was essentially devoid of Ku. Importantly, this preparation of DNA-PKcs was inactive in our assays and was found to contain contaminating Ku at a molar ratio of less than 1 part Ku per 100 parts DNA-PKcs. Notably, although devoid of kinase activity toward a peptide substrate, the preparation of DNA-PKcs lacking Ku still effectively forms adducts with wortmannin that can be competed with ATP (Fig. 4,A). Furthermore, upon addition of purified Ku, no increase in wortmannin binding to DNA-PKcs was observed (Fig. 4 B). These results, therefore, indicate that the ATP binding site of DNA-PKcs is accessible, irrespective of the presence of Ku, and therefore, Ku is likely to control DNA-PKcs activity in some other manner, such as modulating its ability to mediate ATP hydrolysis or by controlling access to the protein substrate.

A defining characteristic of DNA-PK is its ability to be activated by lesions in DNA. To see whether DNA is able to stimulate wortmannin binding to the ATP-binding site of DNA-PKcs, we incubated a preparation of DNA-PKcs/Ku in the presence of increasing amounts of linear double-stranded DNA. Notably, we were unable to see any stimulation of wortmannin binding to the ATP-binding site in these studies (Fig. 4,C). Furthermore, we were also unable to see a competitive effect on wortmannin interacting with DNA-PKcs by adding increasing amounts of a DNA-PK peptide substrate (Fig. 4 D). This suggests that, at least for this peptide, substrate interaction with the DNA-PKcs kinase domain does not modulate binding of wortmannin to key residues in the ATP binding site and suggests that peptide binding occurs to a region distinct from that targeted by wortmannin. On the basis of these results, previous studies on wortmannin interactions with PI 3-kinase, and the sequence homology between DNA-PKcs and the p110 component of PI 3-kinase, the lysine at amino acid residue 3751 of DNA-PKcs is likely to be the site of nucleophilic attack by the furan group of wortmannin.

Recent data have shown that treatment of cells with wortmannin can cause marked sensitization to IR (24, 25, 26, 27). However, it remains unclear as to whether this radiosensitization is mediated through inhibition of DNA-PK. To investigate whether DNA-PKcs is a target for wortmannin in vivo, we incubated HL60 cells in the presence of this molecule. After incubation with the drug for 90 min, nuclear extracts were prepared, and Western blotting was performed using the antiwortmannin antibody. Fig. 5,A clearly shows that a major labeled band that comigrates with wortmannin-labeled DNA-PKcs, is present in wortmannin-treated HL60 nuclear extracts. A similar comigrating band is present in wortmannin-treated HeLa nuclear extract. To establish unequivocally that the wortmannin-labeled polypeptide in these studies is, indeed, DNA-PKcs, we treated the DNA-PKcs-negative cell line MO59J and its DNA-PKcs parental control line, MO59K (29), with wortmannin. Nuclear extracts were then prepared from these cells and analyzed by electrophoresis followed by immunoblotting with the antiwortmannin antiserum. Notably, as shown in Fig. 5 B, such studies reveal a band that comigrates with purified DNA-PKcs that is present in the DNA-PKcs positive line but absent in the DNA-PKcs null line. These data, therefore, reveal that DNA-PKcs is, indeed, a physiological target for wortmannin.

The compounds wortmannin and LY294002, together with their derivatives, have proved to be valuable probes for studying the function of PI 3-kinase in vivo and in vitro. However, with the cloning of additional members of this family of kinases, which share homology but have distinct functions, it has become of key importance to assess the specificity and mechanism of action of these and other inhibitors in more detail. Here, we have investigated the effects of such molecules on DNA-PKcs, a protein serine/threonine kinase of the PI 3-kinase family that plays important roles in DNA repair and DNA damage signaling. We have discovered that LY294002 and the related compounds rutin, quercetin, and quercitrin all inhibit DNA-PK in an ATP-competitive manner, with K_i of 6, 26, 110, and 208 μm, respectively. Furthermore, we have shown that wortmannin inhibits DNA-PK noncompetitively, with a K_i of 120 nm. Notably, these modes of inhibition and relative inhibitory potencies match those shown previously for PI 3-kinase, although DNA-PK is somewhat more refractory to inhibition than PI 3-kinase in each case.

In addition, we have investigated the mode of wortmannin action on DNA-PK in more detail. Thus, we have shown that, as is the case for PI 3-kinase, wortmannin forms covalent adducts with DNA-PKcs in the region of the molecule harboring its kinase domain and that the formation of such adducts is inhibited competitively by ATP. In light of previous work mapping the site of wortmannin binding to PI 3-kinase and the homology between this kinase and DNA-PKcs, it seems likely that wortmannin targets the lysine at amino acid residue 3751 of the DNA-PKcs polypeptide. Having established that wortmannin targets the ATP-binding site of DNA-PKcs, we then moved on to use it as a probe for the status of the DNA-PKcs active site under various circumstances. Previous studies had revealed that DNA-PKcs catalytic function is stimulated greatly by the simultaneous presence of the Ku heterodimer and DNA ends, raising the possibility that this activation mechanism is mediated via conformational changes in the DNA-PKcs catalytic domain to permit ATP binding. However, we found that wortmannin is able to access DNA-PKcs efficiently in the absence of Ku and/or DNA. Thus, although alternative explanations exist, it seems likely that DNA-PK activation does not reflect opening of the ATP-binding site and, instead, is via activating ATP hydrolytic potential and/or access to the peptide substrate. Finally, we used the fact that wortmannin forms covalent adducts with DNA-PKcs to demonstrate that DNA-PKcs is a major target for wortmannin in intact cells.

Previous analyses of the PI 3-kinase family have revealed that it can be roughly divided into two distinct subgroups on the basis of polypeptide size and through characteristic sequence homologies within the PI 3-kinase catalytic domain (17). One of these subgroups comprises the PI 3-kinases and PI 4-kinases, most of which appear to be largely membrane associated or cytoplasmic enzymes involved in intracellular signalling or vesicular trafficking. The other subgroup comprises DNA-PKcs and a series of other large (M_r >250,000) proteins that are involved in cell cycle control, DNA damage signaling, and/or DNA repair. A noteworthy feature of our work is that it broadens the inhibitory profiles of wortmannin, LY294002, and their relatives to a key member of the second subgroup. This is important for several reasons. (a) It reveals that these compounds are not as specific to the PI 3-kinases and PI 4-kinases as previously thought, meaning that studies analyzing their in vivo effects should be interpreted with some caution. (b) It raises the prospect that these small molecule inhibitors also target other members of the DNA-PKcs subgroup of the PI 3-kinase family. In line with this prediction, it has been shown recently that the protein kinase function of two other members of this subgroup, ATM and the human cell cycle checkpoint protein ATR, are inhibited by wortmannin (23, 28). (c) It raises the exciting prospect that derivatives of wortmannin and LY294002 could serve as the basis for the development of more specific inhibitors of DNA-PKcs and its relatives. Given the involvement of DNA-PK and some of its close relatives in DNA repair processes and because it has already been shown that wortmannin and LY294002 radiosensitize mammalian cells in culture (24, 25, 26, 27), an attractive potential application for such inhibitory drugs is the enhancement of the efficacy of cancer chemotherapy and radiotherapy.

We thank members of the Jackson laboratory for their help and encouragement. Thanks also to Matthias Wymann for providing the antiwortmannin antiserum.