Targeting DNA-PK

16.1. Introduction

DNA-PK is a heterotrimeric complex formed in the presence of DNA that is composed of the catalytic subunit of DNA-PK (DNA-PKcs) and the Ku70/80 heterodimer [1, 2]. Ku70/80 is also known as the DNA binding subunit of DNA-PK. DNA-PKcs is one of the largest and most abundant proteins in eukaryotes, spanning 4128 amino acids and weighing ≈ 469 kDa [3]. This protein was discovered in 1985 as a DNA-activated protein kinase in a HeLa extract contaminated with double-stranded DNA [1, 4, 5]. The key observation was that this new DNA-activated protein kinase phosphorylated the alpha isoform of heat shock protein 90 on a SQ/ST motif, which is a known phosphatidylinositol 3-kinase-related kinase (PIKK) substrate motif [6, 7]. Indeed, DNA-PKcs is the largest member of the PIKK family, which otherwise includes ataxia-telangiectasia mutated (ATM), ataxia- and Rad3-related (ATR), and the mammalian target of rapamycin (mTOR) [8].

All PIKKs share a common domain structure, with a kinase domain located in the C-terminal region, flanked by FAT (FRAP, ATM, TRRAP) and PIKK regulatory domains (PRD). At the N-terminus, PIKKs feature alpha helical HEAT (Huntingtin, Elongation factor 3, A subunit of protein phosphatase 2A and TOR1) repeats (Fig. 16.1) [3]. Though the PIKK kinase domain contains motifs similar to those in phosphatidylinositol 3-kinases (PI3Ks, e.g. PIK3CA), PIKKs are serine/threonine kinases which do not phosphorylate lipids.

Fig.16.1.

Graphical depiction of the DNA-PKcs domain architecture

The gene encoding DNA-PKcs (also known as XRCC7) is PRKDC, located on chromosome 8q11 [9]. Phylogenetic studies demonstrate ancient origins of DNA-PKcs with remarkable amino acid sequence conservation among Eukaryota, particularly within the YPRD motif which is located between a phosphorylation cluster (ABCDE) and the FAT domain [3]. The ABCDE cluster contains 6 redundant autophosphorylation sites and together with autophosphorylation sites in another cluster (termed PQR) these enable DNA-PK regulation of V(D)J recombination and DNA damage repair (DDR) by non-homologous end-joining [10–12]. DNA-PKcs is further regulated through phosphorylation by other PIKK family members, including ATM, within the ABCDE cluster and at T3205.

DNA-PK has been implicated in varied biological processes but is best known for its key function in non-homologous end-joining. In this function, DNA-PKcs orchestrates the repair of DNA double strand breaks (DSBs) [13]. In addition to its well-known roles in DNA damage repair, it is increasingly apparent that DNA-PKcs serves roles in regulation of mitosis [14], transcription [15], RNA processing [16], and innate immune response [17]. Although initial preclinical and clinical studies of DNA-PK inhibition have targeted the DDR roles, it is likely that future clinical studies, while continuing to refine the DDR-inhibition strategies, will also be mindful of opportunities to leverage inhibition of other DNA-PK functions.

16.1.1. Insights from SCID Mice and Other DNA-PK Loss-of-Function Phenotypes

Severe combined immunodeficiency (SCID) in humans is characterized by compromised B- and T-cell development and function [18]. The mouse counterpart to human SCID was identified in 1983 by M. Bosma based on the absence of serum immunoglobulins [19]. Subsequent studies demonstrated that the agammaglobulinemia resulted from defective V(D)J recombination, which in turn was caused by DNA-PK definciency due to inactivating mutation in PRKDC [20]. V(D)J recombination is critical for T- and B-cell development and function and requires antigen receptor gene assembly from Variable, Diverse and Joining gene segments. This process is NHEJ-dependent and is initiated by creation of DNA DSBs by recombination activated 1 and 2 (RAG1 and RAG2), which are lymphocyte specific endonucleases [21]. Because this process generates DSBs with hairpin overhangs, these DNA ends need to be end-processed by the endonuclease Artemis before the V(D)J segments can be ligated. Artemis activation is regulated by DNA-PKcs [22, 23]. Because the PRKDC mutation in SCID mice results in loss of DNA-PKcs expression, these mice are characterized by accumulation of hairpin intermediates during V(D)J recombination and absence of functional antigen receptors [24, 25].

Studies in mice have characterized three distinct categories of DNA-PKcs alterations, which are summarized below: (1) complete loss of DNA-PKcs expression, which can result from spontaneously occurring PRKDC mutations in animals [26, 27]; (2) induced loss-of-function mutations in the DNA-PKcs kinase domain; and (3) knock-in mutations preventing(auto)phosphorylation at the ABCDE and PQR clusters.

1. DNA-PKcs null mice demonstrate complete loss of T- and B-cells in line with a SCID phenotype but do not show any other impairment [28, 29].

2. The D3992A substitution, which results in kinase dead (KD) DNA-PKcs is embryonically lethal in mice and results in neuronal apoptosis, similar to that observed in Xrcc4 and LigIV knock-out mice [30, 31]. However, embryonic lethality can be rescued by Ku loss [30]. Furthermore, cells derived from DNA-PKcs KD mice demonstrate greater sensitivity to ionizing radiation than those from DNA-PKcs null mice [30].

3. Alanine substitutions precluding phosphorylation within the DNA-PKcs ABCDE phosphorylation cluster contribute to bone marrow failure and early death in mice [32]. In contrast, mice with alanine substitutions in the PQR phosphorylation cluster develop normally but have moderate sensitivity to ionizing radiation [32].

Interestingly and in contrast to observations in mice and horses, various components of the DNA-PK heterotrimeric complex are essential in human cells. Indeed, all patients with DNA-PKcs mutations reported in the literature have detectable, albeit reduced DNA-PKcs expression, while naturally occurring DNA-PKcs mutations in other animals can result in null phenotypes. The first patient with mutated PRKDC was described by van der Burg and contained a monoallelic L3062R missense mutation within the FAT domain, which did not affect DNA-PK kinase activity but impaired Artemis endonuclease activation [33]. Clinically, this patient demonstrated a SCID phenotype with absence of B and T cells and normal NK cell counts. A patient with PRKDC compound heterozygous mutations had a A3574V substitution (FAT domain) on one allele and an abnormally spliced transcript with loss of exon 16 from the other allele, likely resulting in loss of function [34]. This patient had dysmorphic features and growth failure, microcephaly, seizures, and substantial neurological impairment in addition to the B⁻T⁻NK⁺ phenotype. Two unrelated patients with homozygous DNA-PKcs p.L3062R mutation exhibited defective DSB repair and V(D)J associated with progressive decline in B- and T-cells along with signs of autoimmunity [35, 36]. Interestingly, two siblings with DNA-PKcs p.L3061R mutation both had immune deficiency but differed in the presence of an autoimmune disorder [36]. While it is unclear whether these patients suffered from mono- or biallelic PRKDC mutations, the cases exemplify the variability of symptoms resulting from similar DNA-PKcs mutations.

Several conclusions can be drawn from those observations: (a) the essentiality of DNA-PKcs in humans suggests additional functions compared to non-hominids; (b) the downstream effects of DNA-PKcs mutations depend on which functions they impede; (c) DNA-PKcs appears to have a role in preventing autoimmunity in humans; and (d) the differences observed in animal models must be taken into account when extrapolating DNA-PKcs findings from animals to humans.

16.1.2. DNA-PK Roles in DNA Damage Repair

DNA damage provoked by endogenous or exogeneous mechanisms represents a constant threat to genomic integrity that must be dealt with effectively by intrinsic repair functions. Therefore, DNA damage repair is a key process for genome maintenance and replication fidelity [37]. Upon DNA damage, the DDR system is engaged, recruiting repair factors and activating cell cycle control checkpoints to permit DNA damage repair. DSBs represent the most toxic form of DNA damage, leading to cell death or chromosomal aberrations [38]. In eukaryotes, DSBs can be repaired by several complementary DDR mechanisms. Of those, homologous recombination (HR) and non-homologous end-joining (NHEJ) are the most well studied. The cell’s choice of DDR pathways is context dependent and, in case of HR, restricted to S and G2 phase of the cell cycle because this pathway requires the presence of a sister chromatid to serve as template [39]. In contrast, NHEJ is active throughout the cell cycle but error-prone [40]. NHEJ is also the dominant repair pathway in the G2 phase for ionizing radiation damage distant from the replication fork [41].

DNA-PK is a key factor in NHEJ, consisting of a heterotrimeric complex composed of DNA-PKcs and the Ku70/Ku80 heterodimers, which are also known as the DNA binding subunit. Ku heterodimerization forms a DNA-binding ring which fits around the major and minor DNA grooves and translocates inward upon DNA binding. Because the heterodimer does not make any direct base contacts, it is thought that the Ku-DNA interaction proceeds in a sequence independent manner [42, 43]. Inward translocation of the Ku heterodimer recruits DNA-PKcs to interact with the DNA DSB and form the DNA-PK holoenzyme (Fig. 16.2). Assembly of the complex between adjacent DNA breaks forms a synaptic complex to keep the broken ends in proximity, protecting them from unscheduled processing [44–46]. Depending on the damage encountered, non-ligatable DNA needs to be processed prior to ligation via the XLF-XRCC4-LIGIV complex. This is carried out primarily by the 5′–3′ nuclease Artemis along with other factors such as the 3′-DNA phosphatase/5′-DNA kinase polynucleotide kinase phosphatase (PNKP) [47–49].

Fig.16.2.

Schematic illustration of NHEJ mediated DNA double strand break repair. Ku70/80 and DNA-PKcs form the DNA-PK holoenzyme and recruit downstream effectors that proceed with DNA end-processing and ligation. Created with BioRender.com

The mechanisms by which DNA-PK orchestrates DNA-end processing are incompletely understood, but recent evidence sheds light on how features of the DNA ends influence DNA-PK autophosphorylation and thereby downstream events. Two main phosphorylation clusters have been identified in DNA-PKcs. The ABCDE cluster spanning residues 2609–2647 contains 6 functionally redundant phosphorylation sites that are required for Artemis activation. Phosphorylation within the ABCDE cluster induces a conformational change that releases Artemis from its autoinhibited state and thereby allows for end-processing [50, 51]. Conversely, blocking phosphorylation within the ABCDE region delays DNA-PKcs release from DSBs and impedes end-processing [51]. Hairpin DNA-ends that are generated during V(D)J recombination, a process dependent on NHEJ, promote DNA-PKcs autophosphorylation at the ABCDE cluster. This leads to phosphorylation of the Artemis C-terminal region which is thought to facilitate its de-inhibition [52]. Once end-processing is complete, phosphorylation at the DNA-PKcs PQR cluster limits further processing [12, 53]. In the case of blunt DNA ends or ends with 3′ overhang, the ABCDE cluster protects open DNA and cannot be phosphorylated, which promotes DNA end protection and favors phosphorylation of downstream factors, such as Ku70/80 [54]. This is in line with the observation that hairpinned DNA ends do not activate DNA-PK to phosphorylate TP53 and that TelN restricted DNA, which generates covalently closed DNA ends, leads to autophosphorylation within the ABCDE cluster but fails to activate DNA-PK downstream substrates [51].

Substantial evidence indicates that DNA-PKcs has extensive post-translational modifications, of which phosphorylation is the best studied. In addition to autophosphorylation, the ABCDE cluster can also be phosphorylated by the PIKK family members ATM and ATR, both of which serve key functions in DDR [55, 56]. In fact, it has been shown that ATM can compensate for DNA-PKcs dysfunction, exemplifying the crosstalk among DDR kinases [57]. Other DNA-PKcs post-translational modifications include ubiquitination, PARylation, NEDylation, and acetylation. However, as is the case with phosphorylating events, the biologic impact of these modifications is only very incompletely understood. DNA-PKcs is ubiquitinated and tagged for proteosomal degradation by Ring Finger Protein 144A (RNF144A), which was the first ubiquitinase known to target DNA-PKcs. RNF114A expression is induced by cell exposure to DNA damaging agents, and RNF114A depletion results in DNA-PKcs accumulation and decreased chemosensitivity [58]. Likewise, knock-down of the chaperone protein VCP (valosine containing protein), which binds ubiquitinated DNA-PKcs, results in DNA-PKcs accumulation, elevated DNA-PK activity, and increased DNA damage repair efficiency [59].

ADP-ribosylation by poly (ADP-ribose) polymerases (PARPs) regulates numerous biological processes and PARP inhibitors were the first approved anti-cancer drugs targeting DNA damage response in BRCA1/2 mutated breast cancer. Notably, PARP and DNA-PK can be co-recruited to sites of DNA damage, and PARP proteins can interact with DNA-PK to maintain genomic integrity after DNA DSB induction [60, 61]. PARylation by PARP proteins stimulates DNA-PK activity in vitro and PARP1 knock-down reduces DNA-PKcs expression and activity in nasopharyngeal carcinoma invitro [62, 63]. Conversely, DNA-PK modulates PARP function by phosphorylating PARP in a DNA dependent manner—although the biological impact is poorly understood [64]. Further studies are needed to determine whether cancers with homologous recombination repair deficiency (which are responsive clinically to PARP-inhibition) are hyper-dependent on DNA-PK as a compensatory mechanism for DSB repair. However, the known biologic interactions between PARP and DNA-PK, and the evidence that NHEJ is a compensatory repair mechanism in cells with HRD, provide rationale for exploring therapeutic combination approaches or sequential approaches drawing upon inhibition of PARP and DNA-PK. As discussed later in this chapter, there is also evidence that DNA-PK co-inhibition in cancer cells with homologous recombination repair deficiency can actually impair response to PARP inhibitors. Given the many cross-connections between PARP proteins and DNA-PK, it is likely the clinical benefit, if any, of co-inhibiting these repair kinases will vary greatly in different cancers.

DNA-PK activity is also regulated by crosstalk with nuclear receptors and indeed nuclear receptor signaling can induce DNA double strand breaks and stimulate recruitment of DNA-PK and other DDR factors [65, 66]. In particular, androgen and estrogen receptor signaling regulate transcriptional activity of the PRKDC promoter [67–70]. In addition, DNA-PK can act as a transcriptional co-regulator and phosphorylate various nuclear receptors [71]. These observations raise intriguing questions as to whether DNA-PK signaling roles differ in malignancies with substantial dependence on nuclear receptors.

Many epithelioid caners express epidermal growth factor receptor (EGFR), and high EGFR expression levels have been associated with poor outcomes. Radiation induces EGFR expression and co-treating cells with an EGFR antibody resulted in sensitization to ionizing radiation (IR). These insights had profound clinical impact on how EGFR positive cancers are treated with radiation therapy [72]. Subsequent studies demonstrated that EGFR interacts with DNA-PK and that IR causes EGFR translocation to the nucleus, which then enhances DDR by interaction with DNA-PKcs [73]. Treatment with a monoclonal EGFR antibody inhibits this re-distribution, thereby prevent interaction with DNA-PKcs and explaining why the antibody sensitizes cancer cells to radiation [74–76]. Thus, co-treatment with an anti-EGFR antibody such as cetuximab is now a standard approach to increase sensitivity to radiation therapy in patients.

16.1.3. DNA-PK Roles in Immunity and Autoimmune Disorders

Innate immunity is activated in response to various pathogens, and host detection of cytosolic DNA is a key step in mounting an anti-viral response. Nucleic acids and other pathogen-associated molecular patterns (PAMPs) are sensed by pattern recognition receptors (PRRs), triggering an immune response [77]. Indeed, genomic instability is a major contributor of cytosolic DNA, which itself is potent activator of a type I interferon response [78, 79]. The cGAS-STING pathway is one mechanism that has emerged as a key surveillance system orchestrating antipathogen and anti-tumor immunity [80]. Upon binding to cytosolic DNA, cGAS catalyzes the production of cGAMP, which subsequently activates the stimulator of interferon genes (STING). STING then translocates from the ER to the Golgi, inducing serial phosphorylation events and ultimately activating TBK1 and interferon regulatory factor 3 (IRF3) which results in production of type I interferons. DNA-PK inhibits cGAS by phosphorylation events, accounting for the autoimmune disorders that often accompany DNA-PKcs defects [80]. In addition, DNA-PK can activate IRF3 dependent interferon-1 response independently of cGAS and STING, although the evidence for these roles has been conflicting, depending on the cell types (nonneoplastic vs. neoplastic) and species in which the studies were performed [17, 81]. This is in line with the report of a second, STING-independent DNA sensing pathway in human cells that appears to be undetectable in murine cells [82]. As one example of an apparently cGAS-STING independent role, DNA-PK mediates IRF3 on threonine 135, causing IRF3 nuclear retention and delayed proteolysis in the setting of viral infection [81]. The importance of DNA-PK signaling to activate innate immune responses is further highlighted by studies interrogating infections with the vaccinia virus (VAVC) [83, 84]. These studies demonstrated that the VACV encoded protein C16 binds to the Ku70/80 heterodimer, which blocks DNA-PK-dependent DNA sensing and thereby attenuates innate immune activation.

DNA-PK roles in immunity, like the key DNA-PK roles in DNA damage repair, are an area of active study. It is likely that the intersection of these biologic themes will engender opportunities to enhance both cytotoxicity and immune response by targeting DNA-PKcs in combination therapies for various cancers. Another promising avenue is the role of DNA-PK modulating T-cell tolerance by interaction with the transcription factor autoimmune regulator (AIRE) [85]. DNA-PK phosphorylates AIRE on T68 and S156, thereby regulating AIRE transactivating functions. Consequently, DNA-PK inhibition or loss decreases expression of AIRE target genes.

16.1.4. DNA-PK Inhibition as Therapeutic Strategy

Genomic instability is a hallmark of many cancers and in some cases is attributable to inactivation of DDR proteins that normally serve as guardians of genomic integrity. Well known examples include the mutations and deletions that inactivate homologous recombination repair pathway proteins and which denote vulnerability to PARP inhibitor therapies. Nonetheless, even in cancers with evident genomic instability, other repair pathways have essential roles in preventing the instability (and resultant genotoxicity) from getting entirely out of hand. As discussed above, there is evidence that homologous recombination deficient cancer cells can become hyper-dependent on DNA-PK as an alternate pathway to maintain at least partial capabilities for DSB repair.

For this reason, DNA-PK is an attractive target for anti-cancer therapies. And beyond the possibility of compensatory DNA-PK hyper-dependence in cancers with deficiencies in other DDR pathways, DNA-PK is generally known to limit genotoxic instability induced by DNA damaging chemotherapies. High DNA-PKcs expression levels are accordingly associated with resistance to cytotoxic therapies and thereby associated with worse prognosis [86, 87]. Conversely, DNA-PKcs null mice demonstrate increased sensitivity to DNA damaging therapy, and multiple studies demonstrate that DNA-PKcs inhibition is synthetically lethal in combination with DNA damaging agents (DDAs) [88–90].

Various evidence suggests that mechanisms of cell death resulting from DNA-PK inhibition (DNA-PKi) are influenced by the functional status of cell cycle control. For example, when treating acute myeloid leukemia cells wtih the selective DNA-PKcs inhibitor peposertib, Haines et al. demonstrated that DNA-PKi combined with DNA damaging chemotherapy potentiated compensatory ATM signaling. This led to increased TP53 expression and induction of TP53-dependent apoptosis [87]. In contrast, malignancies with dysfunctional TP53 fail to engage cell cycle checkpoints in response to combinations of DDA with DNA-PKi and enter mitosis prior to completion of DNA damage repair [91]. Such failure of scheduled DSB repair fosters incremental genomic damage, culminating in mitotic catastrophe and apoptotic cell death. TP53 functional status can thus impact cell fate after DNA-PKi, specifically determining the mechanisms of cell death. Interestingly, DNA-PKi monotherapy, although clinically well-tolerated, has very limited efficacy against most solid malignancies. This indicates that NHEJ inhibition alone is insufficient to drive genomic instability to genotoxic levels in vivo [92]. Current clinical evaluations therefore focus on DNA-PKi as a sensitizer towards conventionally dosed DDAs, such as ionizing radiation or topoisomerase II inhibitors, which induce DNA double strand breaks.

Several clinical trials using new-generation DNA-PKcs inhibitors targeting the ATP binding pocket are underway or have been reported upon. In contrast to prior compounds, new-generation small-molecule inhibitors have greater selectivity for DNA-PKcs over PI3K and other PIKK family members [90]. The first in human phase I trial testing the oral DNA-PK inhibitor peposertib (formerly known as M3814), enrolled 31 patients with advanced solid tumors and did not reach the maximum tolerated dose (MTD). Several clinical trials explored peposertib in combination with chemotherapy, e.g. pegylated liposomal doxorubicin (NCT04092270), or radiation therapy (NCT02516813) [93]. The phase I/IIa first in human trial of AZD7648 completed recruitment and will assess AZD7648 as monotherapy and in combination with either pegylated liposomal doxorubicin or olaparib (NCT03907969) [94].

The aforementioned clinical trials of DNA-PKi combined with DNA damaging therapies at conventional dose levels demonstrated a narrow therapeutic index and substantial toxicity [93]. These challenges highlight the need for better tolerated DNA-PKi combination therapies and also for biomarkers that identify cancers particularly dependent on DNA-PK/NHEJ, in which even low doses of DNA-PKi might be active. Notably, the genetic background of various immunodeficient mouse models must be carefully considered when performing DNA-PKi preclinical evaluations. As discussed above, standard SCID mice, which are DNA-PKcs null (DNA-PKcs^−/−) are not informative for DNA-PKi toxicities to nonneoplastic cells although toxicities with DNA damaging agents are heightened in these mice due to the intrinsic DNA damage repair deficiency.

DNA damage repair is a multi-step process with extensive crosstalk among DDR factors, which can elicit compensatory repair pathway activation upon inhibiting specific DDR effectors. Synthetic lethality of PARPi in homologous recombination (HR) deficient cancer is well described and it is possible that HR-deficiency sensitizes some cancers to NHEJ pathway inhibition. Surprisingly, other evidence suggests that DNA-PKi can abrogate the impact of PARPi in HR deficient cancer [95]. In addition to TP53 status, genomic and functional assays interrogating HR-deficiency might therefore prove to be useful in predicting DNA-PKi efficacy.

Effective DNA-PKi combination therapies will likely be defined by further studies of the relationships between DNA-PK and other DSB repair mechanisms—particularly compensatory mechanisms. For example, many deficient cancers are dependent on CDK2 for G2/M cell cycle arrest, and therefore inhibiting CDK2 by targeting the ATR-CHK1-WEE1 pathway can consolidate response to DNA-PKi [96]. Additionally, DNA-PKi synthetic lethality has been observed in ATM defective cancer and likewise ATM signaling can rescue cells from DNA-PKi, providing rationale for co-targeting ATM and DNA-PK.

Altogether, DNA-PK inhibition is emerging as a promising but challenging therapeutic approach in cancer. While primarily targeted for its role in NHEJ DNA damage repair, DNA-PK also regulates other important biologic pathways. These additional roles provide new opportunities to advance cancer treatment but also increase the likelihood of substantial toxicity in the clinic, which underscores the need for compelling and novel rationales that can guide effective clinical translation.