Introduction

In 2010, there were ~22,020 new cases of primary brain and other nervous system cancer, resulting in 13,140 deaths [1]. Of the primary brain tumors, glioblastoma is the most frequent, accounting for ~12–15% of all intracranial neoplasms [2]. The current standard of care for patients diagnosed with glioblastoma is maximal safe surgical resection followed by adjuvant concurrent radiation and temozolomide (TMZ) followed by TMZ alone [3]. Even with this regimen, the prognosis for patients with glioblastoma is poor with a median survival of 14.6 months and 9.8% of patients alive 5 years from diagnosis [3]. Given the dismal prognosis of the disease, there is significant interest in finding new effective therapies to add to the standard treatment regimen.

One potential area for novel therapeutics is in targeting the DNA damage response (DDR) signaling network including DNA repair pathways and the cell cycle. The DDR network is a potentially target rich environment for two main reasons. First, many of the current therapies used in oncology, such as radiation and TMZ, are cytotoxic therapies that damage DNA. Novel agents that target the DDR and modulate the response to DNA damage could potentially sensitize tumors to these therapies. Second, tumors are able to tolerate damage and rearrangements in their genome partially through the loss of DNA damage signaling and checkpoint pathways [4, 5]. Because of this, targeting the DDR might be selectively toxic to tumor cells or reverse therapeutic resistances.

The DDR-DNA repair and the cell cycle

The stability of DNA is of utmost importance to cells in order to survive and proliferate in a controlled manner. Eukaryotic organisms have evolved several complex mechanisms to ensure the both the stability and high fidelity replication of DNA in the face of potential damage. There are numerous types of DNA damage-chemically modified bases, base-pairing mismatches, strand breaks, abasic sites, bulky adducts, cross-links and others [6]. Insults to the integrity of the DNA molecule may be either spontaneous (through deamination or hydrolysis) or result from exposure to various DNA damaging agents. Furthermore, DNA damage generation may be either endogenous (such as oxidative stress from normal cellular processes) or exogenous (such as ionizing radiation (IR), genotoxic chemicals or ultraviolet (UV) light) [7]. Because maintenance of the integrity of the genetic material is of such prime importance, multiple complex pathways have evolved for the surveillance and repair of these various types of DNA damage. These pathways constitute the DNA repair machinery and cell cycle control, processes that are significantly intertwined.

The cell cycle

Cells have a complex system of interacting proteins that govern decisions over growth and proliferation (Fig. 1) [8]. This system incorporates both the intra and extracellular signaling that enable cells to leave a quiescent state and progress through the various phases of the cell cycle [8]. The cell cycle is composed of four distinct phases: gap 1 (G1), synthesis (S), gap 2 (G2) and mitosis (M). Through this process, cells will replicate the over 6 billion bases in the genome and ultimately divide. Before they are ultimately directed to enter the cell cycle, cells may be in a quiescent (G0) phase. In the G1 phase, cells grow and resume biosynthetic processes including the synthesis of enzymes required for S phase. DNA synthesis through high-fidelity replication of each chromosome occurs during S phase, resulting in dual chromatids for each chromosome by the conclusion. S phase is followed by another gap phase, G2, prior to the initiation of M. M results in the symmetric segregation of the chromosomes and division of the cytoplasm to form two daughter cells.

Fig. 1
figure 1

The cell cycle. This figure depicts critical molecules and relationships in the control of the cell cycle. Sites of therapeutic inhibitor development are highlighted

Full size image

Cells will not progress through the cycle until the stability of the DNA molecule is ensured. Following damage to the DNA, the protein kinase signaling molecules ATM and ATR initiate a cascade of events that lead to activation of cell cycle checkpoints and DNA repair (Fig. 1) [9]. While the checkpoint pathways were initially thought to regulate only cell cycle transitions, it is now recognized that signal transduction cascades link DNA damage signaling, checkpoints, and repair [9, 10]. When checkpoints are activated, progression through the cell cycle is halted to allow cells to repair damage. These checkpoints occur at transitions in the cell cycle: the G1/S to ensure adequate cellular size and a suitable environment for replication, the G2/M checkpoint to ensure that DNA has been replicated and damage repaired prior to initiating M, and the spindle checkpoint in metaphase to ensure proper alignment of chromosomes prior to division. If repair fails, cellular death may result from mitotic catastrophe or through p53 dependent or independent checkpoint initiated apoptosis [11]. The G2/M checkpoint may be particularly important as a potential target in cancer therapeutics as many tumors are already deficient in the G1/S checkpoint due to p53 pathway aberrations [12].

DNA repair pathways

The DNA repair machinery is a complex interconnecting set of pathways that respond to the various forms of DNA damage through distinct mechanisms. There are six major DNA repair pathways: base excision repair (BER), nucleotide excision repair (NER), mismatch repair (MMR), homologous recombination (HR)/Fanconi Anemia (FA), non-homologous end joining (NHEJ) and translesion synthesis (TLS) (Figs. 2,3) [4, 7]. DNA damage may also be directly reversed, a mechanism that is thought to be important in responding to damage from alkylating agents such as TMZ (Fig. 4) [13]. Each of these pathways is involved in repair of specific types of DNA damage, though pathways may work together in repairing various lesions. Broadly speaking, DNA damage can be categorized as single strand or double strand damage with double stranded damage regarded as the more lethal lesion. BER, NER and MMR are all involved in the repair of damage to single strands of DNA (Fig. 2) while the HR/FA and NHEJ pathways repair DNA double strand breaks and crosslinks (Fig. 3). The reality is much more complex, with significant cross talk among the various pathways in response to insults to the DNA. Many pathways and associated enzymes are redundant; deficiencies in one pathway or enzyme may be made up by others [4]. For example, damaged bases are removed through the BER pathway, but if the damage is not repaired adequately and an abasic site is present, DNA replication will stall and result in a double strand break, necessitating involvement of a double strand break repair pathway (Fig. 5). This type of cross-talk forms the foundation for the clinically relevant finding of synthetic lethality between poly(ADP-ribose) polymerase (PARP) inhibition (a BER enzyme) and tumors with HR deficiencies [14–16]. Furthermore, pathways may be differentially active depending on the cell type, phase of the cell cycle and specific type of damage [4]. Regardless of the mechanism for repair, DNA damage initiates a common set of signal transduction pathways leading to cell cycle arrest, chromatin structure reorganization and transcription and post-translational modification of DNA repair enzymes [17]. Common elements related to each repair pathway include the recognition and signaling of damage, removal of damage, and repair of the resulting defect.

Fig. 2
figure 2

The BER, MMR, and NER pathways. This figure depicts critical molecules and relationships in BER, MMR and NER. Sites of therapeutic inhibitor development are highlighted

Full size image
Fig. 3
figure 3

The FA, HR and NHEJ pathways and TLS. This figure depicts critical molecules and relationships in FA, HR, NHEJ and TLS as well as a hypothesized interaction with the EGFR/PI3K/AKT signaling pathway. Sites of therapeutic inhibitor development are highlighted

Full size image
Fig. 4
figure 4

TMZ relationship with DNA repair. The O6 (6–8% of events) and N7 (~70%) positions on guanine and the N1 (~15%) and N3 (10%) positions on adenine are methylation sites for TMZ. O6-methylguanine is repaired by direct reversal through the activities of the enzyme MGMT. If damage is unrepaired, the MMR incorrectly pairs a thymidine with the methylated guanine, which, in turn, activates the MMR system to correct the mispairing. This creates a ‘futile cycle’ of MMR that leads to cell death. Intact MGMT is associated with TMZ resistance and mutations in MSH6 lead to acquired resistance to TMZ. N 3-methyladenine is repaired by the BER system. MPG is a DNA glycosylase involved in BER. While most cells have an intact BER pathway, targeting BER through PARP inhibition potentiates the effects of TMZ, especially in MGMT intact cells

Full size image
Fig. 5
figure 5

Synthetic lethality relationships using PARP inhibition. Single base damage, generated from a variety of insults such as oxidative stress or radiation, is normally repaired by the BER pathway. If damage is left unrepaired, replication will be stalled and single base damage will become a double strand break. This double strand break can be repaired normally by HR. Cells that are deficient in HR (such as those with BRCA mutations) cannot repair these breaks efficiently and are thus particularly sensitive to PARP inhibition which generates more unrepaired single base damage. In addition to intrinsic deficiencies in HR, other contexts such as PTEN deficiencies, hypoxia and CDK inhibition are associated with decreased HR and are sensitive to PARP inhibition

Full size image

Base excision repair (BER)

BER is the major pathway that is initiated in response to small areas of DNA damage, including nucleotide residue damage and single stranded DNA breaks, and is the main guardian against damage due to cellular metabolism and reactive oxygen species (ROS) [4, 7, 18]. There are no heritable syndromes associated with defects in BER, which reflects the essential nature of BER in maintaining cellular viability [4, 7].

BER is initiated when a member of the DNA glycosylase family excises the damaged residue (Fig. 3). PARP and polynucleotide kinase (PNK) then facilitate recognition of single strand breaks and abasic sites. Apurinic/apyrimidic endonuclease 1 (APE1) recognizes the abasic site, flips out the deoxyribose residue, cleaves the 5′ side of the DNA backbone and recruits the next enzyme to the site of damage [19]. In short-patch BER (the major BER pathway), the next enzyme is DNA polβ which cleaves the 3′ end with its AP lyase activity followed by filling in of the resultant gap with its polymerase domain. Finally, XRCC1–ligase 3 complex seals the nick [4, 7, 18, 19]. XRCC1 also acts as a scaffold protein and facilitates the entire excision process [7]. In long patch BER, the AP lyase function is unable to completely excise the sugar phosphate due to strand length or complexity and DNA stand displacement occurs. DNA polβ or larger polymerases, polδ and polε, fill in the DNA defect while the damaged strand is still attached at the 3′ end. The PCNA scaffold protein facilitates this reaction. FEN1 endonuclease then cleaves the 3′ end of the damaged strand and the gap is ligated by DNA ligase 1 [7, 19].

Mismatch repair (MMR)

The MMR system is responsible for the correction of mispairing errors made by DNA polymerases and insertions or deletions that result from slippage of polymerases during replication or recombination. There is also interaction with the NER system to recognize and repair DNA adducts [4, 7].

The MMR pathway is initiated when MutSα (a MSH2/MSH6 dimer) or MutSβ (a MSH2/MSH3 dimer) recognizes the mismatched site (Fig. 2). The mismatch is then excised by the binding of MutLα (a MLH1/PMS2 dimer) or MutLβ (a MLH1/MLH3 dimer) to MutSα and subsequent recruitment of the EXO1 exonuclease. The resultant single stranded gap is stabilized by replication protein A (RPA). The gap is finally filled by the actions of DNA polymerase δ (Polδ) and proliferating cell nuclear antigen (PCNA) and sealed by DNA ligase I [20, 21]. An intact MMR system is required for toxicity to TMZ (Fig. 4), which will be discussed below.

Defects in MMR are responsible for microsatellite instability, a hallmark of the ‘mutator phenotype’ seen in many cancers. The Cancer Genome Atlas (TCGA) found that six of the seven hypermutated glioblastoma samples harbored mutations in at least one MMR gene while only one of 84 non-hypermutated samples had such mutations [22]. Germline mutations in genes encoding proteins in this pathway are responsible for hereditary non-polyposis colorectal cancer (HNPCC) and Turcot’s syndrome [4, 7]. While defects in other repair pathways may confer increased sensitivity to DNA damaging agents, defects in MMR confer resistance to DNA damaging chemotherapeutic agents by reducing the ability to sense DNA damage and activate apoptotic pathways [23–25].

Nucleotide excision repair (NER)

NER is the most versatile pathway and is involved in the repair of a number of helix distorting lesions, including adducts induced by DNA damaging chemotherapeutic agents [4, 7]. NER is divided into two subpathways: global genome repair (GGR), a non-specific pathway that surveys the entire genome, and transcription coupled repair (TCR) a process that recognizes damage that obstructs the procession of RNA polymerase II [4, 7]. Inherited defects in the xeroderma pigmentosa genes (XPA–XPG), involved in both pathways, cause xeroderma pigmentosa, a condition that predisposes to UV light induced cancers. Defects in the TCR specific genes CSA and CSB cause Cockayne syndrome, which involves impairment of normal physical and neurological development secondary to lesion-induced apoptosis and premature aging [7]. The difference in the two subpathways is the mechanism for damage recognition. In GGR, protein complexes screen the genome on the basis of distortion, caused by such lesions as cyclobutane pyrimidine dimers (CPD). GGR is inducible in a p53 dependent manner in response to increased cellular stress, especially for repair of CPDs [26]. The stalled RNA polymerase is the signal for TCR, and is removed by CSA and CSB to enable further repair [7, 26].

NER is initiated when XPC recognizes a site of damage and stably binds to HHRAD23B (R23), forming a nidus for other repair protein aggregation (Fig. 2). Proteins such as XPA, RPA, TFIIH and XPG bind to the XPC-R23 heteroduplex and recruit the ERCC1-XPF heterodimeric complex. XPG and ERCC1-XPF have endonuclease activity, which cut the damaged strand DNA at 3′ and 5′ flanking ends, respectively. The resultant 27–30 nucleotide fragment is removed from the DNA strand and the gap is filled by DNA polδ or ε with the aid of PCNA, RPA and RFC. The final nick is sealed by DNA ligase [27].

Double strand break repair-HR and NHEJ

DSBs occur as a result of damage secondary to IR, chemotherapeutic agents, endogenously generated reactive oxygen species, and from the conversion of single strand breaks during replication. They are also produced as an intermediate in such processes as meiosis and immunoglobulin class switching [4, 7, 18, 28]. The repair of DSB occurs either through the actions of HR, which uses complementary DNA from the homologous chromosome as a template to help fill in damaged areas, or the more error prone NHEJ. The preferential pathway for repair is cell cycle dependent—the less error prone HR requires a complementary strand of DNA as a template, a requirement that is only present after DNA replication in S phase.

The first step in the repair of DSBs is damage signaling. The specific mechanism has not been completely elucidated, but an early event following a DSB is the activation of the signaling molecule encoded by the ATM gene (Fig. 3). ATM is a member of the PI3-kinase superfamily and is activated by DNA damage through autophosphorylation, which causes dissociation of the homodimeric complex and exposes the kinase activity at the carboxy terminus [29]. ATM amplifies the damage signal and is involved in the phosphorylation of proteins involved in both cell cycle regulation and DNA repair.

One of the phosphorylation targets of ATM is the RAD50/MRE11/NBS1 (MRN) complex which has exonuclease activity [7]. This complex functions to expose the 3′ ends in the DNA at the site of the DSB in a mechanism that may involve BRCA1 [4]. A RAD52 associated complex may then compete with the KU protein of NHEJ to force the repair down either the HR or NHEJ pathway [28]. The 3′ end then invades a homologous section of the sister chromatid in a reaction dependent on RAD51, BRCA2 and the DNA dependent ATPase RAD54 (Fig. 3) [4, 7, 28]. After a homologous section is found, a DNA polymerase extends the 3′ end past the damage site using the complementary strand and then returns to the original strand followed by ligation by DNA ligase [4, 7]. The resultant complex is known as a ‘Holiday junction’ and is undone by resolvases, at least in meiosis [7, 28].

NHEJ is driven by the actions of DNA protein kinase (DNA-PK), which consists of a regulatory subunit (the KU80/KU70 heterodimer) and a catalytic subunit (DNA-PKcs, a member of the PI-3 kinase superfamily) (Fig. 3). The KU80/KU70 heterodimer binds to double stranded termini in a sequence independent manner and recruits/activates DNA-PKcs [4, 7]. DNA-PKcs binds and stabilizes the DNA termini and, along with the ATM-activated endonuclease Artemis, processes or trims the ends to allow for end joining [18]. The XRCC4/DNA ligase IV complex then anneals the non-complementary ends [7, 18]. This mechanism is highly error prone and only a major pathway in higher organisms due to the presence of an abundance of non-coding sequences [18].

Translesion synthesis (TLS)

TLS is not necessarily a pathway to repair DNA damage, but an alternative mechanism to tolerate damage to the genome. Some lesions such as bulky adducts and thymine dimers cause the normal DNA polymerase, DNA polδ, to stall at a replication fork. TLS involves signaling which allows cells to switch to an alternative, albeit more error prone, polymerase to bypass the sites of mechanical blockage. Cancer cells show dependence on these error prone polymerases [4].

Direct repair

There are multiple mechanisms of direct repair of DNA damage. One of the most immediate mechanisms of direct repair employed by the cell is the proofreading and 3′–5′ exonuclease activity of DNA polδ [30]. Another mechanism is the direct removal of alkyl groups from the O6 position of guanine by the protein product of the gene O-6-methylguanine-DNA methyltransferase (MGMT). O6-alkylguanine DNA adducts can trigger cell death by forcing cells into permanent G2/M arrest or through a futile cycle of MMR (Fig. 4) [31]. The MGMT gene product irreversibly transfers the alkyl group from guanine to it s cysteine-145 residue. Epigenetic silencing of MGMT is a good prognostic factor in patients with glioblastoma who are undergoing chemotherapy with the DNA alkylator TMZ [32].

Targeting the DNA damage response in cancer

While the DDR offers a multitude of possible targets for cancer therapeutics, it is important to remember that the efficacy of anti-cancer therapy requires the establishment of a therapeutic index. If the DDR is modulated in normal cells and cancer by a novel targeted compound to the same degree, then no advantage has been achieved. Only by leveraging differences between cancer and normal cells can we hope to gain a therapeutic index and make true progress. In general, cancer cells have been shown to exhibit greater deficiencies in DDR than normal cells and accumulate more DNA damage [33, 34]. This non-specific observation offers hope that the DDR can be targeted to achieve therapeutic advantage. Furthermore, specific profiling of DDR deficits in cancer cells can further guide the development of targeted therapies.

As monotherapy, DDR targeted agents can take advantage of inherent deficiencies in a cancer cells DDR to achieve synthetic lethality. Synthetic lethality refers to a scenario where mutations in two separate genes are lethal to cells in a setting where either one on its own has a limited phenotype. As mentioned previously, this strategy has been employed for BRCA mutant [14, 15] or HR deficient [16] cells using PARP inhibition (Fig. 5). Research aiming to identify more of these synthetic lethal relationships is ongoing using siRNA or small molecule approaches.

In combination with DNA damaging chemotherapy and/or radiation, novel DDR targeting agents can take advantage of both intrinsic differences in the DDR between normal and cancer cells as well as the variability of response to DNA damaging agents. For example, cells are more or less sensitive to IR depending on the phase of the cell cycle and microenvironment. During S phase, cells are relatively more radioresistant compared with cells in the G1 or G2/M phases. This observation is one of the reasons that radiation is fractionated in order to gain a therapeutic index- to allow cancer cells to ‘reassort’ into more radiosensitive phases prior to another fraction of radiation. But it has also been shown the many proteins involved in HR, a pathway which operates predominantly in the S and G2 phases of the cell cycle, are overexpressed in S phase. HR may then be a target that achieves a therapeutic ratio by both addressing a population that is relatively resistant to radiation, thereby acting synergistically [35]. Furthermore, hypoxic cells are up to three times as resistant to radiation as their more oxygenated counterparts [36]. But hypoxia can also lead to decreased expression of HR proteins [35, 37, 38], creating a subpopulation that, like the BRCA1/2 mutants, might be more sensitive to targeting with PARP inhibition so that the combination of RT and PARP inhibition may be synergistic (Fig. 5) [39, 40].

Targeting the DNA damage response in glioblastoma

Relationship of DNA damage response to current therapies for glioblastoma

Based on the results of the EORTC-NCIC trial, the current standard therapy for newly diagnosed glioblastoma involves the use of two DNA damaging agents-radiation and TMZ [3]. As discussed above, one proposed mechanism of TMZ toxicity is mediated through the methylation of guanine at the O6 position. The methylated guanine then mispairs with thymine rather than cytosine during replication and triggers the DNA MMR system MMR [41]. MMR incorrectly replaces the thymine residue with another thymine, leading to a futile cycle of MMR (Fig. 4). This futile cycle has a variety of possible consequences ranging from senescence to apoptosis or mitotic catastrophe. In fact, it has been shown that an intact MMR pathway is required for toxicity secondary to TMZ [42] and that patients with recurrent tumors following treatment with TMZ frequently harbor mutations in MSH6, leading to resistance and a hypermutation phenotype [25]. The most prominent effect in gliomas is an initial G2/M arrest followed by prolonged senescence in wild type p53 cells and mitotic catastrophe in p53 deficient cells [43]. MGMT can remove the methyl group from DNA prior to initiation of the futile cycle of MMR, leading to a rescue of the cell from TMZ-induced toxicity (Fig. 4) [13]. Patients with tumors that exhibit reduced expression of the MGMT gene, secondary to gene promoter hypermethylation, have a better prognosis than those with tumors expressing the gene [32, 44]. The O6 position of guanine only accounts for ~6–8% of the methylation induced by TMZ, however [45]. The major site of methylation is at the N7 position of guanine, which accounts for ~70% of the methylation events, followed by the N1 (~15%) and N3 (~10) positions of adenine with the N 3-methyladenine and O6-methylguanine felt to be the most lethal events [45–47]. The BER pathway orchestrates repair of N 3-methyladenine while inhibiting BER potentiates the effects of TMZ [48] (Fig. 4). Furthermore, preclinical research has shown that inhibition of BER in glioma cell lines overexpressing N-methylpurine DNA glycosylase (MPG), a DNA glycosylase involved in BER, potentiates the effects of TMZ [49]. In addition, loss of expression of the transcription factor GATA4, a frequent event in GBM, has been associated with TMZ resistance through the regulatory effects on MPG [50]. This mechanism of damage and repair is independent of O6 methylation toxicity and provides an alternate target for modulating the damage from TMZ in resistant tumors [51].

IR causes cytotoxicity mostly through the generation of DNA DSBs [52], which initiate a specific DNA damage response through either HR or NHEJ. Certain heritable conditions resulting from mutations in genes involved in the DSB DNA damage response and repair pathways are associated with increased sensitivity to IR [4, 53–55]. Even though the cytotoxic effects of radiation are more directly related to DSBs, IR induces about 25 single stranded DNA breaks and even more base damaging events for every one double stranded break (DSB) [52]. This suggests that inhibiting the repair of these lesions might be also potentiate IR. Glioblastoma is also thought of as being relatively radiation resistant. Bao et al. [56] showed that glioblastoma tumor initiating cells, as defined by CD 133 expression, were enriched in heterogeneous cell populations in response to treatment with IR. Furthermore, this enrichment was not due to increased expression of CD133 but associated with increased activation of the DDR signaling network and enhanced repair of DNA double strand breaks in comparison with CD133− cells, resulting in decreased rates of apoptosis [56]. The EGFR-PI3K-AKT pathway, a signaling pathway of prime importance in GBM, has also been implicated in the resistance to IR through modulation of DNA DSB repair pathways [57–59]. Strategies targeting the DNA repair pathways could therefore leverage differences in the repair capacities of cancer versus normal cells or serve to potentiate the activity or reverse intrinsic resistance to standard DNA damaging therapies.

Current therapeutic strategies targeting the DNA damage response

Direct reversal of DNA damage

Novel DNA damage response therapies for altering the direct reversal of DNA damage are based on the observation that cells with decreased expression the MGMT gene are more sensitive to alkylating agents [44]. Furthermore, cells that are initially sensitive to TMZ or other monofunctional alkylating agents develop resistance. To reverse this resistance to therapy, a class of DNA repair modulating agents that act as pseudosubstrates for MGMT, such as O6-benzylguanine (O6-BG) and lomeguatrib are being tested in clinical trials (Table 1) [60–64]. These compounds act by mimicking the normal substrates for MGMT, thereby irreversibly inactivating and depleting the enzyme. In one phase II trial, O6-BG restored sensitivity to TMZ in anaplastic astrocytoma, but without a similar effect in glioblastoma [62]. There were also grade 4 hematologic events in 48% of patients, possibly related to the non-specificity of O6-BG for tumors versus normal tissue. This marrow toxicity has limited the value of this class of agents to date.

Table 1 Compounds in current clinical trials that target DNA repair pathways or the cell cycle
Full size table

Base excision repair

Apurinic/apyrimidic endonuclease (APE) overexpression in GBM

The rationale for targeting APE in GBM is related to the observation that in glioma cell lines, APE activities have been found to be increased 10-fold [65]. Furthermore, APE levels are increased in response to oxidative stress in glioma cell lines [66] and elevated APE levels in human tumors are associated with shorter time to progression [67].

Poly(ADP-ribose) polymerase (PARP)

An extremely active area of therapeutic research within the BER pathway is PARP inhibition, either as augmentation of existing therapies or as monotherapy through a synthetic lethal approach. PARP is an enzyme with multiple functions in the cell. One function is to sense single strand breaks and abasic sites as a component of the BER pathway (Fig. 2) [4, 7]. Activation of PARP-1 results in the addition of poly(ADP-ribose) branched chains onto damaged DNA, leading to relaxation of chromatin structure and the recruitment of other proteins involved in the BER pathway [68]. If BER is repressed through PARP inhibition, small lesions that are normally repaired will persist, causing replication fork collapse and conversion into double strand breaks that are repaired by HR (Fig. 5) [14, 15]. Cells that have intrinsically defective HR, either through BRCA mutations [4, 15] or other deficiencies [16], are therefore exquisitely sensitive to BER failure through PARP inhibition. While glioblastoma does not display mutations in classic PARP sensitizing genes such as BRCA1 or BRCA2, any context in which HR might be compromised could provide the conditions for synthetic lethality with PARP inhibitors. One example is the hypoxic fraction of cells where HR protein expression is decreased [35], a subpopulation that might be more sensitive to targeting with PARP inhibition based on the relationship between HR and BER [39, 40]. This targeting of the hypoxic cell fraction, a population that has been is resistant to radiation, could potentiate the effects of radiation and act synergistically. Another example is the high rate of PTEN aberration in GBM. PTEN deficient glioblastomas have been shown to harbor defects in HR and exhibit sensitivity to monotherapy with PARP inhibitors through synthetic lethality [87, 88]. While not strictly as synthetic lethal phenomenon, EGFRvIII mutant glioma cell lines may also be sensitive to PARP inhibition due to increased levels of ROS [89]. HR deficiency can also be induced. Recently, investigators have shown synthetic lethality by simultaneously modulating both BER and HR pathways. CDK1, aside from its role in the cell cycle, is also a kinase that promotes BRCA1 and HR activity. Inhibition of CDK1 has been shown to sensitize cells to PARP inhibition (Fig. 5) [69].

Inhibiting BER has implications not only for chemotherapy, but for radiation as well; as described above, radiation produces more than 25 times the amount of single strand breaks and base damage as double strand breaks [52]. Conversion of this base damage into the more lethal double strand breaks would be expected to enhance toxicity. PARP has also been implicated in double strand break repair, either through influencing the balance of HR and NHEJ [70] or by participation in alternative NHEJ pathways [71]. Synthetic lethal approaches may also be useful in combination with other therapies in a context-specific manner. Pre-clinical models have shown that PARP inhibition sensitizes cells to both radiation [72–78] and various chemotherapeutic agents [72, 73, 75, 79] and clinical trials are currently underway (Table 1) with some studies already completed [80, 81].

PARP inhibition has been studied in glioma cell lines in combination with radiation and/or TMZ. In a study investigating the radiosensitizing effects of the PARP inhibitor KU-0059436 on four distinct glioma cell lines, abrogation of PARP activity was associated with an increased sensitivity to radiation [82]. The sensitizing effect of KU-0059436 was limited to replicating cells and enhanced in cells that were in S-phase. This specificity could potentially be important in providing a therapeutic advantage in clinical use, as the normal brain cells are not actively replicating. Cells treated with the PARP inhibitors were shown to exhibit an increase in DSBs that arose during DNA replication and DNA repair by HR was increased in the S-phase cells. The authors hypothesized that this increase in DSBs was related to replication fork collapse. Given the increased sensitivity in S-phase, the radiosensitizing effect of KU-0059436 was enhanced by fractionated radiation rather than single dose, thereby allowing cells to reassort in the cell cycle to a more sensitive phase [82]. This S-phase specificity was also shown in HeLa cells [77].

PARP inhibition also sensitizes glioblastoma cells to the effects of TMZ and reverses the TMZ resistance in cell lines and xenografts [51, 79, 83]. In TMZ sensitive cell lines, the addition of the PARP inhibitor CEP-6800 to TMZ significantly increased the magnitude of DNA damage, the length of time required for repair, and the number of cells arrested at the G2/M boundary [79]. These cellular changes were associated with enhanced tumor sensitivity in glioblastoma xenografts as well [79]. PARP inhibition can also restore sensitivity to TMZ in resistant tumors [51, 83]. As described previously, cells that are deficient in MMR are resistant to TMZ [42, 84]. While TMZ toxicity is thought to be related to methylation of the O6 position of guanine [85], the majority of alkylation events occurs on the N3 and N7 positions and PARP-1 is important in the repair of methylation at the N7 position [47]. In glioblastoma xenograft models that were resistant to TMZ due to deficient MMR, treatment with the PARP inhibitor INO-1001 resensitized tumors [51]. Russo et al. [86] investigated the addition of the PARP inhibitor E7016 to RT and TMZ in a U251 xenograft model. Investigators randomized mice bearing human glioblastoma xenografts to vehicle, E7016, TMZ/RT or E7016/TMZ/RT. Absolute growth delay was 6.6 days for E7016 alone, 20.2 days for TMZ/RT and 31 days for the E7016/TMZ/RT, showing that PARP inhibition in addition to the current standard combination resulted in improved outcomes. Interestingly, in the in vitro portion of this study, E7016 sensitized cells to radiation alone without showing an increase in the quantity of DSBs. Instead, a persistence of DSBs led the authors to hypothesize that the mechanism was related to an inhibition of DSB repair rather than SSB repair and collapsed replication forks [86].

There are currently four ongoing phase I and II trials with PARP inhibitors in patients with glioblastoma, either in the newly diagnosed or recurrent setting, in combination with TMZ alone or with TMZ and radiation therapy (Table 1) [90].

Double-strand break repair

Association with PI3K-AKT pathway

A connection between the PI3K-AKT pathway and DNA repair was suggested after it was observed that PTEN deficient mouse embryonic cells exhibited genomic instability related to RAD51 mediated DSB repair or cell cycle checkpoint deficiencies [91]. But the interaction between DDR and EGFR/PI3K/AKT is likely extremely complex with broad intersection between pathways [92]. For example, while PTEN deficient cells have been shown to harbor deficits in HR and a decreased capacity for repair through this mechanism, activation of the PI3K pathway has been shown to be a poor clinical prognostic factor in glioblastoma, possibly through an association with radioresistance [93]. Common mutations in glioblastomas that have an association with PI3K-AKT and radioresistance include PTEN and EGFR [94, 95]. PTEN deficient glioblastoma cell lines have high basal level of AKT activation and exhibit radioresistance [57]. When AKT activation in PTEN deficient glioblastoma cells is reversed through induction of PTEN or direct inhibition of PI3K, radiosensitivity is conferred through delayed repair of DNA DSBs [57].

The EGFRvIII mutant, a common mutation in glioblastoma, contains a large deletion in the extracellular domain leading to constitutive activation [96–98]. Expression of EGFRvIII in glioma cell lines and orthotopic models results in enhanced repair of radiation induced DSBs [58, 99, 100]. EGFRvIII likely promotes radioresistance through the NHEJ pathway as inhibition of DNA-PKcs abrogates the radioresistance of EGFRvIII mutant cell lines [58]. These data suggest that targeting the PI3K-AKT pathway or NHEJ may help reverse the radioresistance seen in many forms of glioblastoma. Clinical trials of erlotinib and gefitinib in combination with radiation and TMZ for patients with glioblastoma have thus far been disappointing [101–103], but this is possibly due to a lack of efficacy in patients with glioblastoma specific alterations.

c-Abl

RAD51 is an essential component of the HR pathway. It has been shown that the phosporylation of RAD51 in response to radiation is dependent on c-Abl [104] and that the fusion tyrosine kinase BCR/ABL promotes drug resistance through induction of HR [105]. Researchers at the NCI used imatanib mesylate (Gleevec) to induce radiosensitivity in glioma cell lines through the inhibition of the ABL kinase [106]. Inhibition of ABL led to a reduction in the formation in RAD51 foci in response to ionizing radiation. Importantly, this reduction in RAD51 foci and associated radiosensitivity in response to imatanib mesylate was not seen in normal fibroblasts, raising the possibility of a therapeutic window. Imatanib mesylate in combination with radiation was shown to increase growth delay and improve survival in murine glioblastoma models [107]. While imatanib mesylate did not show a clinical benefit in glioblastoma patients as monotherapy [108], a phase I/II trial is studying the effects of combination therapy with radiation (Table 1) [109].

Proteasome inhibition

While proteasome inhibitors would be expected to have multiple effects within a cell, Murakawa et al. [110] showed that inhibition of the proteasome interfered with RAD51 formation and inhibited HR without effecting NHEJ. Proteasome inhibition prevents the disassembly of MDC1 foci leading to ineffective recruitment of BRCA1 [111]. Furthermore, proteasome function is required for the activation of the FA pathway. Proteasome inhibition results in the inhibition of FANCD2 monoubiquitination, prevents radiation induced FANCD2, BRCA1 and RAD51 foci formation, and results in the persistence of DNA damage [112]. Bortezomib, a proteasome inhibitor used in the treatment of hematologic malignancies, has been shown to be safe when given with radiation to the brain and TMZ in a phase I trial and is currently being studied in multiple phase I and II trials alone or in combination with other therapies for patients with glioblastoma (Table 1) [113].

Cell cycle

G2/M checkpoint abrogating agents

Many tumor cells are deficient in the G1 checkpoint of the cell cycle and rely more heavily on the G2 checkpoint in response to DNA damage [114]. By blocking the G2 checkpoint, cells enter M prematurely before DNA repair has been completed in response to damage, leading to cell death [115]. CDK1/CDC2 is the primary molecule that governs the G2/M checkpoint and entry of cells into M (Fig. 1) [116]. The cell cycle is able to progress into M when CDK1/CDC2 is in its dephosphorylated state. The G2/M checkpoint is thought to be activated by multiple mechanisms, depending on what phase of the cell cycle that DNA damage occurs [117]. Damage during the G1 and S phases results in a signal cascade that primarily relies on the actions of ATR (Fig. 1). Once recruited to sites of DNA damage and activated, ATR exerts its influence over the checkpoint through phosphorylation of CHK1 at Ser-317 and Ser-345 [118, 119]. Phosphorylation of CHK1 activates its own kinase activity, leading to phosphorylation of CDC25C, which results in its sequestration in the cytoplasm [116]. CDC25C is then not able to activate CDK1/CDC2 through its phosphatase activity, thereby halting cellular progression into M. CDK1/CDC2 is also regulated by the WEE1 and MYT kinases, which block progression into M through phosphorylation of the Tyr15/Thr14 residues (Fig. 1) [120–122]. Numerous compounds that inhibit CHK1, thereby abrogating the G2 checkpoint, have shown efficacy as sensitizers in preclinical [123–133] studies and have been investigated clinically in phase I [134–138] and phase II [139] trials (Table 1). Inhibition of WEE1 or MYT kinases through RNA interference or small molecule inhibitors have also shown the ability to abrogate the G2/M checkpoint and sensitize cells to DNA damage [119, 140].

G2/M checkpoint abrogating agents in glioblastoma

Damage induced by TMZ normally leads to arrest at the G2/M checkpoint and repair of DNA. The TMZ mediated G2/M arrest was abrogated upon administration of the CHK1 inhibitor UCN-01 in one study using human glioblastoma cell lines [126]. UCN-01 administration reduced levels of phosphorylated CDC25C and CDK1/CDC2. Upon removal of UCN-01, cells exhibited a transient increase in phosphorylated CDK1/CDC2 and G2/M arrest. This abrogation of the G2/M checkpoint was associated with a decrease in prolonged senescence in wild-type p53 and enhanced mitotic catastrophe in both wild-type and p53 deficient cells. CHK1 inhibition also increased the cytotoxicity of TMZ in a p53 independent manner [126].

WEE1 inhibition has also been thought to be a particularly attractive target in glioblastoma [141]. In silico analysis of gene expression microarray data found that WEE1 was the most frequently overexpressed kinase in glioblastoma [142]. Fernet et al. [143] examined the effect of WEE1, ATM and PARP inhibition on the cellular response to irradiation in glioblastoma cell lines. The investigators found that cellular accumulation in G2 following irradiation in the G1 or S phase required WEE1 kinase activity. WEE1 therefore represents another potential target for sensitizing glioblastoma cells to radiation therapy or other DNA damaging agents that result in G2/M accumulation.

Potential limitations of DNA damage response modulation

While the excitement associated with many DNA damage response modulators is certainly warranted, there are a number of potential problems with employing them for therapeutic gain. When used in combination with conventional chemotherapy and radiation, DNA repair inhibitors would be expected to increase the toxicity of therapy as well as the efficacy. Aside from situations of synthetic lethality, where a context- or cancer-specific deficit is leveraged, DNA repair modulators that increase the damage caused by conventional agents may do so without expanding the therapeutic advantage. Increasing the damage induced by radiation in a non-specific manner would be no different than simply increasing the radiation dose. For this reason, Phase I trials and/or safety run-ins for combination therapy will be essential even though these compounds might be considered ‘targeted.’ In addition to concerns about toxicity, the promise of DNA damage response modulation faces practical hurdles. First, many DNA repair enzymes may not be ‘druggable,’ that is, there may not be sites on the proteins for which it is easy to engineer potent inhibitors. DNA damage responses may also be tumor-specific, necessitating alternative therapeutic strategies and limiting the generalizability of results. Furthermore, even if DNA damage is enhanced in tumor cells, anti-apoptotic mechanisms may dominate the tumoricidal responses to DNA damage, thereby reducing efficacy. Hypothetically, DNA repair inhibitors could even increase the aggressiveness of tumors by promoting error-prone mechanisms of repair, thereby increasing mutation frequency and genomic instability. Secondary mutations have already been shown to lead to resistance to both PARP inhibitors and cisplatin [144, 145]. Taken together, these limitations suggest that there is still much work to do to realize the potential of DNA damage response modulators to achieve significant therapeutic gains.

Conclusions

Targeting of the DNA repair machinery and cell cycle control offers an exciting realm for future novel therapeutic development. Tumors frequently exhibit defects in the DDR signaling pathways, offering a potential weakness that can be exploited through rational complementary targeting in order to provide a therapeutic advantage. Furthermore, inhibiting critical DNA repair and cell cycle pathways can enhance the damage that is already delivered by current therapies. If this enhanced damage can be increased in tumor cells versus normal cells given potential intrinsic differences in DDR, therapeutic gain could be achieved. Pre-clinical models have shown compelling results and early phase trials are underway.