The Challenge of Combining Chemo- and Radiotherapy with Checkpoint Kinase Inhibitors

The majority of chemotherapeutics, as well as radiotherapy, either directly damage DNA or target basic cellular and metabolic processes that lead to indirect DNA damage. The DNA damage response (DDR) elicits activation of cell-cycle checkpoints to promote repair by pausing the cell cycle or, in cases of unrepairable DNA damage, stimulate programmed cell death (1, 2). As such, much effort has been directed to augment the efficacy of chemotherapeutics and radiotherapy through combination with cell-cycle checkpoint kinase inhibitors as this forces cells to progress through the cell cycle in the presence of unrepairable DNA damage, ultimately leading to cell death (3). Hence, regimens of combined chemotherapy or radiotherapy with checkpoint kinase inhibitors have proven beneficial in several preclinical studies and the biology and mechanisms behind these combinations have been extensively reviewed (4–6). These preclinical observations have led to the assessment of many of these combination therapies in clinical trials in oncology and, as such, this article provides a perspective on the challenges of reliably translating the robust preclinical findings into clinical benefit.

The majority of cancer cells harbor genomic alterations, most commonly (50%–70%) in the form of a dysfunctional G₁ checkpoint, due to loss of function of p53, p21, or p-Rb, greatly enhancing their dependency on the remaining checkpoints for the repair of DNA damage (6–9). Therefore, combining chemotherapeutics or radiotherapy with compounds targeting (intact) intra-S and G₂–M checkpoint kinases are predicted to have favorable clinical outcomes based on the rationale that tumor cells lacking any functional checkpoints will not be able to arrest the cell cycle and repair DNA damage (reviewed in refs. 3, 6, 10, and 11). This irreversible damage results in replicative catastrophe or is taken through into mitosis, where tumor cells undergo programmed cell death (6, 12). Several druggable targets have been identified that can be inhibited to disrupt the intra- S and G₂–M checkpoints. In the presence of replication stress or DNA damage, cyclin-dependent kinase (CDK) activity is restrained by different checkpoint kinases, including ATR and ATM, WEE1, and checkpoint kinases 1 (CHK1) and 2 (CHK2) (reviewed in refs. 13, 14; Fig. 1, Box 1).

Activity of the CDK1–cyclin B complex, which becomes more active through S-phase and reaches peak activity at the G₂–M transition (Fig. 2), is directly restricted by phosphorylation events through WEE1/MYT1, which catalyzes inhibitory tyrosine-15 phosphorylation of CDK1/2. ATR and CHK1 can restrict CDK activity by inhibiting the CDC25 family of phosphatases, responsible for dephosphorylation of tyrosine 15 (21). Hence, control over the balance of CDC25-WEE1 regulation is essential for establishing and maintaining both the intra-S and G₂–M checkpoints (14).

Targeted cell death by combination therapies, resulting from elevated CDK activity, occurs through two main pathways: replicative and mitotic catastrophe, although they are not necessarily mutually exclusive. First, ATR, CHK1, and WEE1 play an important role in regulating CDK activity within the S-phase checkpoint (18, 22). Inhibition of these checkpoint kinases can lead to higher CDK activity in S-phase and elevated replication origin firing (15), thereby increasing the cellular demand for dNTPs. Moreover, ATR, CHK1, and WEE1 are critical for deoxyribonucleotide triphosphate (dNTP) production in response to replication stress (23) through promoting the synthesis (24), activation, and stabilization (25, 26), of ribonucleotide reductase. Therefore, their inactivation can lead to under-replication, and depending on the mutational background, replicative catastrophe (18, 27–29). Second, regions of under-replicated DNA could be taken through into mitosis, where mitotic catastrophe is induced through chromosome missegregation or delayed cell death after micronucleation (reviewed in ref. 30). Inhibiting the DDR through abrogation of the damage induced G₂–M checkpoint, either via CDC25A stabilization by CHK1 or ATR inhibition or via CDK1 re-activation through WEE1 inhibition, poses additional cytotoxicity (11, 31). As cells subjected to G₂–M checkpoint inhibition cannot downregulate CDK1 activity, they enter mitosis and attempt to perform chromosome segregation with extensive DNA damage, ultimately leading mitotic catastrophe and cell death (Fig. 2).

Even though some chemotherapeutic agents have the ability to kill cells throughout the cell cycle, targeting all proliferating cells, some commonly used chemotherapeutic agents display cell cycle phase–specific toxicity, exerting a lethal effect that is maximal in a specific stage of the cell cycle (Fig. 3; refs. 32, 33). It is useful to distinguish between agents that generate actual DNA damage and therefore induce the DDR (e.g., radiotherapy, which leads to production of free oxygen radicals in the cell, or chemotherapy drugs, such as etoposide) from those that primarily act by blocking replisome activity. The replisome is the molecular machinery that performs parental strand separation and DNA synthesis and its inhibition therefore increases replication stress and targets S-phase cells specifically. For example, while replisome poisons initially induce stalled replication forks, and therefore activate the ATR–CHK1 axis, these blocks can ultimately result in DNA damage if the fork is not protected or eventually processed into a DNA double-strand break (DSB), thereby inducing other checkpoint kinases, including ATM and CHK2. As such, targeting specific checkpoint kinases requires an in depth understanding of their roles within the DNA damage and replication stress checkpoint response. For example, ATR and WEE1 inhibition exhibit different modes of action (Fig. 1; ref. 34). In contrast to CHK1 and ATR, WEE1 is not only central to the replication stress response, but also functions as a core regulator of CDK activity (14). Mechanistically, it is also important to note that screens identified CDC25A inactivation as the top gene “hit” conferring resistance to ATR and CHK1 inhibitors (35, 36), while the main determinants for WEE1 inhibitor resistance included CDK2, SKP2, and CUL1 inactivation (37), all required for timely entry into, or progression through, S-phase. Furthermore, as ATR functions in the very early stages of the DDR pathway, alternative methods to activate CHK1 do exist (38). This suggests there might be some differential response to ATR, CHK1, and WEE1 inhibitors in the clinic, depending on the genetic background of the patient's cancer. Therefore, selecting appropriate checkpoint inhibitors (ATM, ATR, CHK1/2, or WEE1) that target one stage of the cell cycle in combination with agents that are maximally cytotoxic during that stage of the cell cycle provides the rationale for combination and multiple rounds of treatment.

In this perspective, we identify four key areas for improvement in the translation of preclinical observations of combination therapies into rationally designed clinically trials: (i) Selection of drug combination partners – clinical trials need to take into account the phases of the cell cycle where anticancer drugs exert their lethal and blocking effects as identified in preclinical studies; (ii) Sequence of drug administration – as cells that do not harbor any DNA damage are not dependent on checkpoint kinases for survival, DNA damaging agents have to be administered at an appropriate time prior to checkpoint kinase inhibition, to achieve the desired antitumor effect; (iii) Appropriate dose reductions – preclinical data shows that in the context of combination therapy, less than the MTD of anticancer therapeutics produces therapeutic benefit. Translating this into the clinical setting could reduce levels of toxicity to the normal tissue; and (d) Patient stratification – patient stratification is essential for effective treatment, based on their genetic alterations and biological insights obtained during preclinical and phase I studies.

Here, we provide a perspective on preclinical studies testing “cyclotherapy” (reviewed in ref. 39), defined as a strategy that exploits loss of checkpoint function, through use of inhibitors or due to genetic defects, in combination with DNA-damaging agents. We argue that in combination therapy, the selection of DNA-damaging agents should not be based upon the cancer subtype, but rather on cell-cycle specificity and the genetic background of the tumor. Furthermore, we briefly describe pharmacodynamic biomarkers that can be used to monitor the success of such combination regimens. We hope that this perspective will shed new light on the rational design of cyclotherapy trials.

The combination of intra-S and G₂–M checkpoint kinase inhibitors and genotoxic agents has been extensively studied in preclinical models to determine the most effective drug combinations. In Fig. 3, we classified chemotherapy agents based on their action during certain phases of the cell cycle and several preclinical studies have demonstrated the importance of this concept when choosing drug combination partners. For instance, Montano and colleagues showed that SCH9000776 (CHK1i) sensitizes cells to hydroxyurea (20- to 70-fold) and gemcitabine (5- to 10-fold), but not to cisplatin or SN-38 (active metabolite of irinotecan), a topoisomerase I inhibitor, in a number of cell lines (40). SN-38 induces cell-cycle arrest when the replication fork collides with an inhibited topoisomerase I, resulting in DNA DSBs that subsequently activate both Chk1 and Chk2 (41, 42). Even though Chk1 is required for homologous recombination (HR)-mediated repair of these DSBs, ATM-Chk2 is essential to the G₂–M arrest, and hence, predominates over the control of cellular survival in this context (41). It is not surprising, therefore, that concurrent incubation of SN-38 with the ATM inhibitor KU55933 caused a dramatic sensitization (40). Similarly, Chk2, but not Chk1, depletion dramatically increased the proportion of apoptotic HCT116 cells exposed to camptothecin (CPT), another topoisomerase I inhibitor (41). Even though the clinical use of CPT was hindered due to severe toxicity and low solubility and stability issues, many other semisynthetic CPT derivatives, including irinotecan and topotecan, continue to be important anticancer agents (43). Therefore, the relevance of cell-cycle phase specific cytotoxicity, and the relative importance of the canonical ATR-Chk1 and ATM-Chk2 pathways in each of those stages, needs to be considered in the design of combination therapies.

Not only the choice of drugs, but also the sequencing and schedule of chemotherapy agents and cell -cycle checkpoint inhibitors impacts on the effectiveness of combination therapy. In many cases, delayed, but not concurrent, treatment with checkpoint inhibitors sensitizes cancer cells to S-phase–specific chemotherapeutic agents (44). For example, the combination of CHK1i with gemcitabine or hydroxyurea (HU) is only effective when CHK1 inhibition occurs >18 hours after treatment with these chemotherapy agents (Fig. 3) as shown by experiments in a number of cell lines, in vivo patient-derived xenografts, and ex vivo studies (40, 45, 46). Gemcitabine and HU are both examples of clinically approved inhibitors of ribonucleotide reductase (RR), which catalyzes the rate-limiting step of the formation of dNTPs, the precursors of DNA synthesis, by direct reduction of the corresponding ribonucleotides (47, 48). Hence, through inhibition of RR, gemcitabine and HU limit synthesis of dNTPs such that replication slows or stops, and therefore cause an accumulation of S-phase–arrested cells only after prolonged treatment, where RPA-coated replication forks become more dependent on Chk1 for stability (40, 46). Therefore, as chemotherapeutic agents cause replication to slow down over increasing time, cells become more sensitive to inhibition of Chk1 (44, 49). Similar results have been found in preclinical experiments using NCI-H2122 and NCI-H441 non–small cell lung cancer cell lines and in vitro/in vivo models with combination therapy of CHK1i (LY2603618) and pemetrexed, which disrupts the production of folate, essential for DNA replication. The activity of LY2603618 manifested only after pemetrexed had caused maximum S-phase arrest, which required prolonged drug exposure (50). Furthermore, cells that had already completed DNA synthesis were not likely to be sensitive to the combination of these agents, and as such, it takes time for an entire asynchronous population of cells to traverse S-phase before an effect was observed, and multiple rounds of treatment might therefore be beneficial (50).

Combinations of checkpoint kinase inhibitors and chemotherapy agents will be additively toxic and as such, doses should be reduced appropriately when drugs are used in combination (32, 51). The objective should be to identify a combination schedule that effectively inhibits the target [i.e., the minimum biologically effective dose (BED); ref. 52] rather than the MTD for the checkpoint inhibitor, in combination with the appropriately timed induction of the required DNA damage by the combination partner (cytotoxic or radiation induced). This is particularly important in terms of developing a schedule for combination therapies, where tolerability over time is important to maximize clinical benefit (51). If the drug regimen is designed rationally, and phase-specific agents are added at the cell-cycle phase where they can exert their maximal effect, one may not need to use the MTD of each individual drug to exert an antitumor effect, thereby limiting toxicity-related issues (51). One of the important aspects of the translation of preclinical data into the clinic is the selection of a starting dose for in-human trials (53), which is complex for single-agent trials, but even more so for combination therapies. Detailed analysis and understanding of both human and mouse pharmacokinetics for the agents concerned as well as their interactions are required, and discussion of these complex issues is beyond the scope of this review.

Preclinical studies of CHK1i in combination with HU in MDA-MB-231 breast cancer cells, revealed that the addition of CHK1 inhibitors SCH900776 (1.25 μmol/L) or UCN-01 (200 nmol/L) caused a 60-fold decrease in the IC₅₀ for HU (40), and these results were further validated in a number of other cell lines. Similar reductions in IC₅₀ values were found for gemcitabine and cytarabine, which interferes with DNA synthesis, in MCF10A cells, when these drugs were combined with SCH900776 (CHK1i; ref. 40).

In many cases, continuous exposure (>24 hours) to a checkpoint kinase inhibitor or a chemotherapy agent alone, such as CHK1 inhibitors (MK-8776 or LY2606368), cisplatin or the WEE1 inhibitor (AZD1775) at high concentrations (IC₈₀) is insufficient to robustly induce γH2AX, which provides a read-out of endogenous DNA damage and causes subsequent cytotoxic effects (49, 54, 55). Combination treatments, however, including CHK1 inhibition and cisplatin, CHK1 inhibition and radiation, and third, WEE1 inhibition (AZD1775, also known as adavosertib) and irinotecan, demonstrate greater-than-additive induction of γH2AX in the majority of cells (45%–75%; refs. 55–57) and hence, the concentrations of these agents may be greatly reduced when used together.

As discussed previously, intra-S and G₂–M cell-cycle checkpoint inhibitors are especially effective when administered in the context of a dysfunctional G₁–S checkpoint, due to the inactivation of endogenous inhibitors (p16 or p21 family) or p53-mediated changes, or alterations causing premature S-phase entry (CCNE1 amplification, RB1 loss, or CDKN2A mRNA downregulation; refs. 5, 58). For example, the WEE1 inhibitor (WEE1i) MK-1775 (AZD1775) was shown to preferentially kill p53-deficient tumors by using p53 matched-pair cell lines (54). As the G₁ checkpoint is compromised due to TP53 loss and these cancer cells thus rely solely on the G₂–M checkpoint after DNA damage, G₂–M checkpoint abrogation through small-molecule inhibition of WEE1 sensitizes p53-deficient cells to DNA-damaging agents, including gemcitabine and platinum compounds (59). However, there is opposing literature as to whether WEE1 inhibition is more effective in p53 mutant backgrounds compared with p53 wild-type cancer cell lines (59–61).

Emerging data suggest that cancers with increased intrinsic replication stress (e.g., oncogene-induced, that is, arising from HPV E6/E7 and/or PARK2/FBXW7 loss-dependent cyclin E1 dysregulation) and/or attenuated DNA damage repair pathways, may exhibit higher sensitivity to cell-cycle checkpoint kinase inhibitors, such as prexaseritib (CHK1i) and berzosertib (ATRi; refs. 5, 58, 62–64). This emphasizes that there are alternative ways of reducing checkpoint function in cancer, and hence, we need to improve on our methods to call an individual tumor's checkpoint control deficiencies to improve patient stratification. Therefore, pretreatment tissues from patients should be analyzed by targeted next-generation sequencing and correlated with response.

Furthermore, predictive pharmacodynamic biomarkers, used to verify target inhibition and subsequent modulation of the downstream signaling pathways in vivo, should be monitored throughout treatment and are crucial in predicting the response (65). Biomarker levels depend on the type of chemotherapeutic agent, the concentration of the drug, and the duration of time between combination treatment and analysis of the biomarker, and could help make informed decisions about treatment continuation and optimal scheduling. A number of IHC markers, including γH2AX (a marker of DNA damage; ref. 66), p-H3 (which correlates with chromosome condensation in mitosis), and p-CDC2 (a marker of CDK inhibition), can be used to assess effectiveness across the range of combination therapies discussed here (67, 68). Recent studies have shown that increased DNA damage is responsible for a synergistic antitumor effect and correlates closely with potentiation of two drugs, as cell lines that were not sensitive did not show any additive changes in γH2AX (54). As such, γH2AX levels can help guide treatment timing as exemplified in Montano and colleagues (49). Addition of CHK1 inhibitor (MK-8776) following gemcitabine in xenografts at 42 hours led to only a 10% increase in cells staining positive for γH2AX, whereas CHK1i treatment 18 hours post-gemcitabine led to a 50% increase, suggesting that at later time-points following gemcitabine, cells have already recovered from gemcitabine-mediated replicative stress (49).

In Table 1, we present a comprehensive survey of current clinical trials testing combinations of checkpoint inhibitors, specifically WEE1, ATM, ATR, CHK1, and CHK2, together with DNA-damaging agents (https://www.clinicaltrials.gov). Recent analysis of phase II clinical trials showed that over half of the sampled phase II studies did not cite any matched preclinical or preceding clinical evidence, suggesting that the design of clinical trials adding cell-cycle checkpoint inhibitors on top of standard of care chemo- or radiotherapy does not always rest on a sound mechanistic and biological basis (69). Here, leading on from the discussion on preclinical data, we focus on scheduling, dosing, and patient stratification in clinical trials and the lessons that can be learned from failed and successful combination therapy trials. The data we have collated demonstrate that the clinical efficacy of combining checkpoint kinase inhibitors with cytotoxic agents is influenced by the drug regimen used.

As previously discussed, preclinical studies have demonstrated that the delayed administration of cell-cycle checkpoint kinase inhibitors after S-phase–specific chemotherapeutic agents is important for maximal cell killing. Even though this delayed drug schedule has been adopted in many of the clinical trials presented in Table 1, and results have been encouraging, other studies have shown limited efficacy and could have benefited from timed sequential therapy, a therapeutic strategy that exploits cell-cycle kinetics by administering a second drug in close approximation with the damage and/or replication stress induced by the first drug.

One of the most prominent examples of recent success comes from a phase II trial of gemcitabine and berzosertib, a selective ATR inhibitor, in high-grade serous ovarian cancer (NCT02595892), where it was shown that combination therapy increased progression-free survival compared with gemcitabine alone [mPFS was 22·9 weeks (17·9–72·0) for gemcitabine + ATRi and 14·7 weeks (90% confidence interval (CI) 9·7–36·7) for gemcitabine alone (HR, 0·57; 90% CI 0·33–0·98; one-sided log-rank test P = 0·044; ref. 63)]. In this study, the ATR inhibitor, berzosertib, was administered 24 hours after gemcitabine, highlighting the enhanced clinical efficacy when the optimal sequence and delayed schedule of administration of checkpoint kinase inhibitors in combination with chemotherapy is applied in the clinic. Similarly, a phase II study (NCT01357161) assessing the effectiveness of AZD1775 plus paclitaxel and carboplatin (P/C) inTP53-mutant ovarian cancer with an analogous dosing strategy (AZD1775 or placebo bid for 2.5 days plus paclitaxel (175 mg/m2; i.v.) and carboplatin (AUC5; i.v., on day 1) also found that AZD1775/P/C was associated with a significant increase in progression-free survival (PFS) when compared with P/C alone (70).

Those using concomitant treatment [gemcitabine and AZD7762 (ATRi; NCT00413686)], or a short (30 minutes) cell-cycle checkpoint inhibitor bolus 24 hours postchemotherapy [gemcitabine and CHK1i (NCT00839332)] have, on the contrary, shown very limited improved efficacy over standard of care chemotherapy. This could be due to inadequate target engagement, as the above-mentioned studies used continuous intravenous administration of the cell-cycle checkpoint kinase inhibitor (>1 hour, up to 1–3 days), or the ability of cells to recover from checkpoint inhibition. A phase I study tried to evaluate the efficacy of a 3- versus 24-hour PF-00477736 (CHK1i)-intravenous infusion 24 hours after gemcitabine treatment in Advanced Solid Tumors (NCT00437203), which would have provided an interesting insight into the importance of timing of administration and target engagement, but unfortunately, the study was terminated prematurely due to business reasons.

A phase I study combining AZD1775 (WEE1 inhibitor) and carboplatin in platinum-resistant ovarian cancer (NCT02272790), showed an objective response rate (ORR) of 67% and a median progression-free survival (mPFS) of 12.0 months when carboplatin was administered on day 1 and AZD1775 on days 1–3, 8–10, and 15–17 of 21 day cycles. When the same drugs were used in a different regimen (carboplatin was administered on day 1 and AZD1775 only on days 1–3), mPFS was 4.2 months, with an ORR of only 30%. This indicates the importance of continued target inhibition (i.e., WEE1) in the presence of DNA damage and suggests that repeated target engagement is necessary to block the ability of cells to recover from checkpoint inhibition and thus for the therapy to be effective.

Below, we consider two examples of trials that have been unsuccessful and discuss these in the context of the preclinical findings presented in the previous section. First, a phase I/II study designed to compare the OS of LY2603618/gemcitabine to gemcitabine alone in locally advanced or metastatic pancreatic cancer (NCT00839332), as well as another study testing gemcitabine and CHK1i on a similar schedule in Japanese patients (NCT00937664, terminated), did not only reveal a higher number of serious adverse effects, but also failed to observe any objective responses with concomitant administration (71). Preclinical studies have shown that concurrent CHK1 inhibition and gemcitabine failed to activate DNA helicases, which function to unwind the DNA in the absence of nucleotides, ultimately resulting in DNA cleavage and replication catastrophe (29, 46). These two clinical studies demonstrate that sensitization of this drug combination in patients does not occur with concurrent treatment, which aligns with the preclinical findings.

A final example is that of AZD1775 (WEE1i) and gemcitabine (NCT02272790) in TP53-mutated epithelial ovarian cancer, where AZD1775 was administered orally on days 1 and 2 and gemcitabine intravenously on day 1 (week 1–3). The mPFS in this group was 1.7 months, versus 5.5 months (when AZD1775 was combined with paclitaxel), and 12.0 months (when AZD1775 was combined with carboplatin; ref. 72). The current consensus is that WEE1 does not play a direct role in replication fork protection, but rather that its effects are mediated more via its roles in replication dynamics and indirect DNA DSB generation (14). This emphasizes the need to select the agent that gives a maximally cytotoxic effect when used in combination with WEE1 inhibitors, in this case carboplatin or paclitaxel over gemcitabine, and the need to test this in preclinical models or early-phase clinical trials before starting a phase II study.

A high number of clinical studies had to be terminated due to serious adverse effects (SAE). In most cases, toxicity is dose-dependent, as with a phase I dose-escalation study of cisplatin and UCN-01 (CHK1i), where cisplatin was infused over 1 hour before UCN-01 (45 mg/m²/d) given as a 72-hour continuous infusion. Escalation of cisplatin was planned through five dose levels at 20, 30, 45, 60, and 75 mg/m², where accrual was halted at dose level 2 due to dose-limiting toxicities consisting of grade 5 sepsis with respiratory failure (73). A number of other examples can be found, where after a planned safety analysis of the first-line trial, dose and schedule were reduced and modified, respectively, to reduce toxicity (NCT02087241, carboplatin/pemetrexed + AZD1775 and NCT02087176, docetaxel + AZD1775).

Molecular diagnosis and treatment stratification are at the heart of precision medicine in cancer. While the increasing availability of genomic and proteomic data has led to some key advances, molecular stratification becomes much more complex when targeting intracellular processes with multiple control mechanisms and feedback loops. Discovery and validation of cell-cycle (arrest) biomarkers has been challenging, especially as distinct predictive genes emerge from different treatment cohorts and cancer subtypes. Dysregulation of cell-cycle control can be assessed in multiple ways, but the most commonly used predictive biomarker for G₁-S checkpoint deficiency is TP53 mutation status.

A phase I study evaluated AZD1775 (WEE1i) as monotherapy or in combination with chemotherapy in patients with refractory solid tumors (NCT00648648), and they correlated the response with TP53 status (74). Their study suggested that tumors from responding patients were mildly enriched for TP53 mutations, given the response rates of 21% and 12% in TP53-mutated and TP53 wild-type patients, respectively (74). The modest difference in response rate between TP53 wild-type and TP53-mutant patients was not sufficiently strong to promote the use of TP53 as a biomarker for selection of patients for Wee1 inhibitor therapy, but multiple clinical trials were, however, built on this premise. NCT00031681, NCT02157792, NCT03641313, NCT03641547, NCT02906059, NCT02448329, NCT01849146, NCT02272790, NCT01164995, NCT02087176, NCT01357161, NCT02508246, NCT02101775, NCT02585973 and other studies have subsequently implemented patient stratification (based on TP53 mutation or others) and/or retrospective mutation correlation analysis in their clinical trial design.

Some improvement on single gene (i.e., TP53) testing has been implemented. Recently, a clinical trial was launched to study AZD1775 and other drugs (as a monotherapy, nonetheless), based on a biomarker-driven treatment strategy (NCT02688894), in patients with small-cell lung cancer where patients were allocated to different treatment arms depending on the presence or absence underlying mutations in cell-cycle–related genes (CDNK2 and TP53) and the MYC family genes (75). Similarly, the FOCUS4 programme (ISRCTN90061546), consisting of multiple biomarker-stratified randomized trials, enrolls patients in different arms of the trial, depending on their molecular background. Those with TP53 + RAS mutations, a selection marker for loss of G₁–S control and oncogene-driven replication stress, receive AZD1775 as a monotherapy (76). Furthermore, recent studies have suggested that more functional biomarkers of replication stress, including surrogate read-outs of the presence of ssDNA (e.g., phospho-RPA), could indicate a higher dependency on the ATR/CHK1/WEE1 pathway of the replication stress response (77). Recently, a novel transcriptomic signature of replication stress was developed, which could predict response to both ATR (P < 0.018) and WEE1 inhibitor (P < 0.029) treatment in both cell lines and human pancreatic cancer organoids. This signature takes into account expression of numerous genes involved in cell-cycle control and DNA maintenance (e.g., MYC, CCNE1, E2F1, KRAS, and CCND2) and has potential as a putative biomarker in a clinical setting (64).

Despite the limited observed clinical efficacy of WEE1 and ATR inhibitors as monotherapies in some genetic backgrounds, there might be patients, especially those with DDR-defects, that do show a partial response to monotherapy checkpoint kinase inhibitor treatment (78). For instance, the ATR inhibitor BAY 1895344 showed stronger antitumor activity in patients with solid tumors that harbor ATM loss (78). Similarly, CRISPR screens and bioinformatic analyses have recently been performed to determine which genetic alterations render cancer cells significantly more sensitive or resistant to checkpoint kinase inhibitors as a monotherapy (64, 79). Such screening has been performed for ATRi in a TP53-mutant background providing insight into which genetic background will show increased sensitivity or resistance and it is anticipated that many such screens will follow (79). However, a recent study in pancreatic cancer patient–derived cell lines shows that deficiency in DDR pathways only renders sensitivity to checkpoint kinase inhibitors when cells also suffer from intrinsic replication stress (64). Therefore, tumors that are DDR-deficient may benefit more from treatment with cytotoxic agents, such as platinum-based therapies, or PARP inhibitors, which can be combined with ATR or WEE1 inhibition if concurrent high replication stress exists within the tumor (64). These studies demonstrate the challenges that come with using checkpoint kinase inhibitors as a monotherapy.

Finally, novel biomarker approaches using other cell-cycle inhibitor(s) or combinations should be further investigated. We note that to improve the monitoring of patient responses, the development of a broad spectrum of blood-based biomarkers, such as those described here, will also be important in order to avoid further patient biopsies where possible.

Cyclotherapy approaches attempt to optimize drug selectivity against tumor cells with defective G₁–S control, and this proves to be a successful therapeutic strategy if dosage and timing strategies are carefully considered in the design of clinical studies. We are suggesting that the most important consideration in combination therapy is cell-cycle specificity of the cytotoxic agent and the checkpoint kinase inhibitor. This may mean that novel combinations of cytotoxic agents with checkpoint kinase inhibitors may be more fruitful than previously approved cytotoxic agents. A number of genetic alterations or signatures of elevated replication stress could be utilized to identify tumors that are predicted to display higher sensitivity to checkpoint kinase inhibitors.

There is evidence to justify delaying administration of CHK1 or WEE1 inhibitors after DNA-damaging agents to achieve a more significant decrease in tumor growth, because such a schedule would allow cancer cells to become increasingly dependent on intra-S and G₂–M checkpoint signaling for HR-mediated repair of exogenously induced replication stress and DNA damage. We propose the efficacy of combination therapies and treatment schedules should be assessed in a first cycle of drug administration for routine clinical management, which would then be continued if the predicted pharmacodynamic biomarkers are observed.

Ideally, patient treatment should be stratified according to the genetic background of their cancer as different mutations will exhibit varying degrees of sensitivity to these agents compared to normal tissue. Enhancing the therapeutic window is clinically relevant as it will reduce toxicity to the healthy cells while maintaining therapeutic benefit. Patient stratification can be achieved through predictive or preclinical testing methods, or through retrospective clinical correlation methods.

In addition, retrospective analysis of patient genomes can help guide further clinical studies. To explore the effect of the mutations mentioned above on the response rate, as well as identify other genetic markers that predict sensitivity, including mutations in genes involved in DNA repair (such as BRCA1), cell-cycle regulation (including cyclins) and oncogene-induced replication stress genes (such as c-MYC and KRAS), clinical trials with increasing patient numbers should be conducted.

A final consideration returns to fundamental science, namely, the specificity of the inhibitors. The WEE1 inhibitor, AZD1775, has been shown to inhibit PLK1 and the CHK1 inhibitor, prexaseritib (LY2606368), as well as other CHK1 inhibitors, often carry varying levels of CHK2 inhibitory activity (80). This fundamental issue with the specificity of an anticancer drug highlights the necessary and crucial interdependence of preclinical and clinical scientists, as emphasized throughout this article, to ensure that the best science is applied reliably and robustly to the life-critical field of oncology.

T.S. Maughan reports grants, personal fees, and nonfinancial support from AstraZeneca and grants from Merck-Serono outside the submitted work, as well as grant funding to institution from Psioxus and DMEC support from Pierre Fabre. No disclosures were reported by the other authors.