Development and Evolution of DNA-Dependent Protein Kinase Inhibitors toward Cancer Therapy

DNA double-strand break (DSB) is considered the most deleterious type of DNA damage, which is generated by ionizing radiation (IR) and a subset of anticancer drugs. DNA-dependent protein kinase (DNA-PK), which is composed of a DNA-PK catalytic subunit (DNA-PKcs) and Ku80-Ku70 heterodimer, acts as the molecular sensor for DSB and plays a pivotal role in DSB repair through non-homologous end joining (NHEJ). Cells deficient for DNA-PKcs show hypersensitivity to IR and several DNA-damaging agents. Cellular sensitivity to IR and DNA-damaging agents can be augmented by the inhibition of DNA-PK. A number of small molecules that inhibit DNA-PK have been developed. Here, the development and evolution of inhibitors targeting DNA-PK for cancer therapy is reviewed. Significant parts of the inhibitors were developed based on the structural similarity of DNA-PK to phosphatidylinositol 3-kinases (PI3Ks) and PI3K-related kinases (PIKKs), including Ataxia-telangiectasia mutated (ATM). Some of DNA-PK inhibitors, e.g., NU7026 and NU7441, have been used extensively in the studies for cellular function of DNA-PK. Recently developed inhibitors, e.g., M3814 and AZD7648, are in clinical trials and on the way to be utilized in cancer therapy in combination with radiotherapy and chemotherapy.


Introduction
Ionizing radiation (IR) is thought to exert a variety of biological effects through the induction of damages on DNA. One Gy of X-ray or γ-ray is estimated to induce approximately 500 thymine glycols, 150 DNA-protein crosslinks, 1000 single-strand breaks, and 40 double-strand breaks (DSBs) [1]. DSB is considered the most deleterious among the various types of DNA damage.
In eukaryotes, DSB is repaired mainly through homologous recombination (HR) and non-homologous end joining (NHEJ) [2]. There are two other pathways, i.e., alternative end joining (A-EJ) and single-strand annealing (SSA) [2]. A-EJ is also termed microhomologymediated end joining (MMEJ) or DNA polymerase theta-mediated end joining (TMEJ). These four pathways are distinguished by their usage of sequence homology ( Figure 1). HR reconstitutes the DNA sequence around DSB using a homologous or identical sequence as the template, which is usually longer than 100 base pairs (bp). On the other hand, NHEJ utilizes little or no sequence homology, i.e., 0-4 bp. A-EJ and SSA utilize sequence homology of 2-20 bp and more than 50 bp, respectively. HR, A-EJ and SSA are preceded by the end resection, which creates single-stranded DNA with 3 -overhang. The end resection proceeds in two stages, i.e., initial short-range resection (≈100 nucleotides (nt)) followed by long-range resection (several hundred or thousand nt). While A-EJ requires only short-range resection, HR and SSA require long-range resection. In NHEJ, the DNA ends that are not compatible for ligation undergo end processing, which results in the deletion or insertion of nucleotides at the junction. In addition, joining of the ends in close vicinity may sometimes lead to ligation of incorrect pairs of DNA ends, resulting in chromosomal aberrations such as deletions, inversions, and translocations. Thus, NHEJ is considered more error-prone than HR. However, HR in vertebrates has a requirement for the sister chromatid and is restricted to late S and G2 phases. (Note: a very recent study demonstrated that DSBs at the centromere are repaired through HR even in the G1 phase [3].) The majority of cells are in G1 and G0 phases, in which cells rely on NHEJ to repair DSBs. In human cells, NHEJ accounts for approximately 80% of DSB repair even in the G2 phase [2]. Moreover, in most cases, the deletion or insertion of a small number of nucleotides can be tolerated, because only a small portion of the genome encodes proteins. A-EJ and SSA are thought to be more error-prone than NHEJ, because they are apt to occur between repetitive sequences, resulting in the loss of the sequence in between.
NHEJ is also implicated in the process of V(D)J recombination in vertebrate immune system [2]. Enormous diversity of immunoglobulins and T cell receptors are generated through the recombination of V (variable), D (diversity), and J (joining) segments, each of which can be selected from a number of segments. Recombination activating gene 1 and 2 (RAG1 and RAG2) induce a cleavage between the selected segments and the flanking recombination signal sequences. Then, the segments are joined through NHEJ. Thus, NHEJ is thought to be of prominent importance especially in vertebrates such as humans.

DNA-PK and Its Role NHEJ
DNA-dependent protein kinase (DNA-PK) is composed of a DNA-PK catalytic subunit (DNA-PKcs) and Ku heterodimer (hereafter denoted Ku), which consists of Ku80 (also termed Ku86) and Ku70 [4,5]. DNA-PK binds to and is activated by the end of a doublestranded DNA (dsDNA). Thus, DNA-PK acts as the sensor for the end of dsDNA, which appears when a DSB is generated.
The kinase activity of DNA-PKcs is required for NHEJ because the catalytically inactive (kinase-dead) form of DNA-PKcs cannot rescue the radiosensitivity and V(D)J recombination defects of DNA-PKcs-deficient cells [57,58]. Although the precise roles of protein phosphorylation by DNA-PKcs remain elusive, DNA-PKcs is shown to phosphorylate NHEJ factors and other potentially NHEJ-related proteins ( Table 1). The significance of phosphorylation of each substrate protein has been discussed elsewhere [59].

Regulation of chromosome segregation
A number of small molecules that inhibit DNA-PK have been developed to date. These compounds have been powerful tools to delineate the function of DNA-PK. Furthermore, they are promising agents in cancer therapy, sensitizing cancer cells to radiotherapy and chemotherapy. Hereafter, the development of DNA-PK inhibitors and their potential in cancer therapy are reviewed.

Development and Evolution of DNA-PK Inhibitors-Pursuit for Potency and Selectivity
Since the discovery of the importance of DNA-PK in DSB repair through NHEJ, a number of small molecules inhibiting DNA-PK were developed from the 1990s to 2000s. Most of the inhibitors were developed on the basis of the structural similarity of DNA-PK to PI3K. In this phase, high potency, i.e., low 50% inhibiting concentration (IC 50 ), and selectivity, i.e., high IC 50 for other kinases especially PIKKs and PI3Ks were pursued. Some of the products, such as NU7026 and NU7441, were useful tools for the functional studies on DNA-PK.

OK-1035
The first reported inhibitor OK-1035, 3-cyano-5-(4-pyridyl)-6-hydrazonomethyl-2pyridone ( Figure 3) was found after screening over 10,000 natural and synthetic compounds [61]. The IC 50 for DNA-PK activity in vitro was initially reported to be 8 µM [61] but was later reported to be 100 µM [62] (Table 2). OK-1035 retarded the DNA repair in cultured murine leukemia cells at 2 mM [63]. Although the IC 50 of OK-1035 was at least 100-fold higher for other kinases such as Protein Kinase C, the effects of OK-1035 on PI3Ks and PIKKs were not tested. OK-1035 suppressed the accumulation of p53 and the induction of p21 in response to adriamycin treatment, suggesting that it might have inhibited ATM and/or ATR as well [64].
By screening a library of LY294002 derivatives, NU7026, 2-(morpholin-4-yl)-benzo[h] chromen4-one ( Figure 5B), was found [69,88]. The IC 50 of NU7026 for DNA-PK was 230 nM, which was much lower than that of ATM, ATR, mTOR and PI3K [69,88] (Table 2). The morpholine ring structure appeared essential for inhibitory activity. NU7026 at 10 µM sensitized cultured cells to radiation in a manner dependent on DNA-PK [69,88] (Table 3). NU7026 was also shown to potentiate the cytotoxicity of topoisomerase II poisons [89]. A preclinical pharmacokinetics study was conducted, showing rapid plasma clearance of NU7026 through metabolism [90]. Recent studies demonstrated that NU7026 administered intraperitoneally (i.p.) at 25-50 mg/kg could potentiate the tumor growth suppression via radiation and chemotherapeutic drugs salinomycin and TRAIL-inducing compound 10 (TIC10) in vivo, i.e., in xenograft in immunodeficient mice [91][92][93] (Table 3).  Figure 5C), showed more potent inhibition of DNA-PK than NU7026 [70,94]. The IC 50 of NU7441 for DNA-PK was 14 nM [70,94] (Table 2). The latest structural study by cryo-EM showed the insertion of the chromen and morpholine groups into the deepest hydrophobic pocket of DNA-PKcs formed by Leu3751, Tyr3791, Ile3803, Leu3986 and Ile3940 and the insertion of the dibenzothiophene group into another hydrophobic pocket formed by Met3729, Pro3735 and Leu3751 [39]. These multiple interactions between NU7441 and DNA-PKcs would explain the higher affinity and selectivity of NU7441 than wortmannin for DNA-PKcs. To date, NU7441 has been most frequently used in functional studies of DNA-PK. NU7441 sensitized cultured cells to IR and etoposide in a manner dependent on DNA-PKcs at 0.5 µM [95] ( Table 3). NU7441, 10-25 mg/kg, i.p., could potentiate tumor growth suppression by radiation and chemotherapeutic drugs in vivo [95,96] (Table 3). Figure 5D), was developed by the modification of NU7441 to increase water solubility [68,97]. KU-0060648 exhibited an IC 50 of 5 nM for DNA-PK, which is still lower than NU7441 but also inhibited PI3Ks at lower concentrations [68] ( Table 2). Hence, KU-0060648 acts as a dual inhibitor for DNA-PK and PI3Ks. Growth inhibition was observed above 30 nM in cultured cancer cell lines and above 10 mg/kg in tumor xenografts [97][98][99] (Table 3). Sensitization to chemotherapeutic drugs was observed in similar dose ranges [97][98][99] (Table 3). Figure 5E), in which the chromenone structure in NU7441 was replaced by benzoxazinone, was identified as a more selective inhibitor for DNA-PK [72]. While IC 50 for DNA-PK was comparable to or higher than NU7441, IC 50 for PI3Ks was more than two orders of magnitude higher [72] (Table 2). LTURM34 was shown to restore partial chemosensitivity to chemoresistant prostate cancer cells at 3 µM [100] (Table 3).

Development and Evolution of DNA-PK Inhibitors-Pursuit for Clinical Availability
The potent and selective inhibitors developed above may have been anticipated for applications in cancer therapy, but this was not feasible due to pharmacokinetics and toxicity. In the 2010s, additional inhibitors were developed and are now under clinical trials. Some of them are not selective inhibitors for DNA-PK and are even more inhibitory to mTOR and/or PI3Ks. There are also selective inhibitors for DNA-PK, which are expected to be used in combination with radiation and chemotherapeutic agents. In general, the inhibitors for clinical use show increased solubility in water for oral availability. They also tend to have a large structure and be inserted deeply into the ATP-binding pocket, being in contact as well with amino acids which are not conserved among PIKKs or not in contact with ATP ( Figure 2D). Thus, increased contact will enhance the potency and/or selectivity of these inhibitors.

LY3023414 (Samotolisib)
Phase 1 and 2 clinical trials are in progress, and the results of three monotherapeutic phase 1 studies [132][133][134] and one phase 2 study [135] have been reported to date. There is a report of a phase 1 study combining LY30234014 with Notch inhibitor crenigacestat (LY3039478). In these trials, the inhibition of PI3K and/or mTOR rather than DNA-PK may be expected. Three monotherapeutic phase 1 studies have shown tolerable safety properties with a recommended phase 2 dose (RP2D) of 200 mg administered twice daily (BID) [78,124,125]. In the phase 2 study, recruiting cancer patients with activating PI3K mutations showed only modest clinical activity [135]. The combination of crenigacestat and LY3023414 exhibited poor tolerance, which resulted in lowering the dose and reduced clinical activity [136].
In a phase 1 study, M3814 was well tolerated with RP2D of 400 mg BID [143]. Concentrationdependent reduction in DNA-PKcs autophosphorylation in peripheral blood mononuclear cells was also observed [143]. Although M3814 monotherapy did not show partial response in this study [143], the combination with radiotherapy or chemotherapy is currently under phase 1 clinical trial, and the results are anticipated.
Phase 1 and 2 clinical trials of AZD7648 monotherapy and its combination with doxorubicin or olaparib are currently in progress.

Summary and Future Directions
As seen above, a number of DNA-PK inhibitors have been developed. Potent and selective inhibitors in the earlier generation, such as NU7026 and NU7441, have been useful tools in functional studies of DNA-PK. In the next generation, dual inhibitors such as LY3023414 and CC-115, and DNA-PK-selective inhibitors such as VX-984, M3814 and AZD7648, have promising effects in preclinical studies and are now under clinical trials. The development of these inhibitors in the 1990s and 2000s was greatly promoted by the systematic chemical modification of previously identified PI3K inhibitors to increase the potency and selectivity toward DNA-PK. On the other hand, M3814 and AZD7648 were found recently through large-scale de novo screening, underscoring the importance of this approach. The latest study elucidating the structure of the complex of DNA-PKcs and inhibitors indicated a possibility of structure-guided drug development.
Dual inhibitors LY3023414 and CC-115 have exhibited growth inhibitory effects in monotherapy as seen in preclinical studies. On the other hand, VX-984, M3814 and AZD7648 showed modest growth inhibition at most but potent sensitization to IR and DNA-damaging chemotherapeutic agents. Thus, the growth inhibition by dual inhibitor may be due primarily to the inhibition of PI3K/mTOR. In order to utilize the potential of dual inhibitors in inhibiting DNA-PK, future studies on the combination with radiotherapy and/or chemotherapy are anticipated.
Recent preclinical studies of M3814 and AZD7648 indicate promising combinations. M3814 showed enhancement of the efficacy of chemoradiotherapy and radioimmunother-apy. AZD7648 potentiated the effects of olaparib, especially to a great extent in ATMdeficient cells. There are lines of evidence indicating that the carriers of pathogenic ATM mutation, accounting for 1-2% in human populations, exhibit a several-fold increased risk of breast cancer [147,148]. Since ATM plays a pleiotropic role in the maintenance of the genome, elevated cancer risk may be caused by haploinsufficiency or the second hit, i.e., the loss of the active allele. It is assumed in the latter case that cancer cells have lost ATM function, whereas normal cells retain it. The cancer cells would then be selectively sensitized to the combination of olaparib and DNA-PK inhibitor. Further studies may be warranted to explore the effects of other combinations and in other genetic statuses.