A phase 1b study evaluating the safety and preliminary efficacy of berzosertib in combination with gemcitabine in patients with advanced non-small cell lung cancer

Objectives: Berzosertib (formerly M6620, VX-970) is an intravenous, highly potent and selective, first-in-class ataxia telangiectasia and Rad3-related (ATR) protein kinase inhibitor. We assessed the safety, tolerability, preliminary efficacy, and pharmacokinetics (PK) of berzosertib plus gemcitabine in an expansion cohort of patients with advanced non-small cell lung cancer (NSCLC). The association of efficacy with TP53 status and other tumor markers was also explored. Materials and methods: Adult patients with advanced histologically confirmed NSCLC received berzosertib 210 mg/m 2 (days 2 and 9) and gemcitabine 1000 mg/m 2 (days 1 and 8) at the recommended phase 2 dose established in the dose escalation part of the study. Results: Thirty-eight patients received at least one dose of study treatment. The most common treatment- emergent adverse events were fatigue (55.3%), anemia and nausea Gemcitabine had no apparent effect on the PK of berzosertib. The objective response rate (ORR) was 10.5% (4/38, 90% confidence interval [CI]: 3.7 22.5%). In the exploratory analysis, the ORR was 30.0% (3/10, 90% CI: 9.0 61.0%) in patients with high loss of heterozygosity (LOH) and 11.0% (1/9, 90% CI: 1.0 43.0%) in patients with low LOH. The ORR was 33.0% (2/6, 90% CI: 73.0%) in patients with high tumor mutational burden (TMB), 12.5% (2/16, 90% CI: 2.0 34.0%) in patients with intermediate TMB, and 0% (0/3, 90% CI: 0.0 – 53.6%) in patients with low TMB. Conclusions: Berzosertib plus gemcitabine was well tolerated in patients with advanced, pre-treated NSCLC. Based on the observed clinical efficacy, future clinical trials should involve genomically selected patients.

Conclusions: Berzosertib plus gemcitabine was well tolerated in patients with advanced, pre-treated NSCLC. Based on the observed clinical efficacy, future clinical trials should involve genomically selected patients.
Ataxia-telangiectasia-mutated (ATM) and Rad3-related protein kinases (ATR) play critical roles in the DNA-damage response (DDR) by regulating the cell cycle checkpoint control and repairing damaged DNA by homologous recombination [8]. In response to DNA replication stress, induced or exacerbated by chemotherapies such as gemcitabine, ATR is recruited to regions of exposed single-stranded DNA to mediate replication fork stabilization, whereas ATM responds to DNA doublestrand breaks [9].
Berzosertib (formerly M6620, VX-970) is an intravenous (IV), highly potent, and selective first-in-class inhibitor of ATR [10]. In preclinical studies, berzosertib sensitizes lung cancer cells to DNA-damageinducing chemotherapeutics such as gemcitabine [10]. Previous clinical studies have shown that berzosertib in combination with chemotherapy is well tolerated with preliminary efficacy signals in several solid tumors [11,12]. Furthermore, a recent proof-of-concept phase 2 study evaluating the berzosertib-topotecan combination reported an objective response rate of 36% (9/25), with a median duration of response of 6.4 months, in patients with SCLC, including those with platinum-resistant disease [13].
Berzosertib efficacy can be enriched in the presence of specific tumor genetic alterations. Tumor protein p53 (TP53) mutational status has been shown preclinically to correlate with response to DNA-damaging agents combined with ATR inhibition [14]. This is explained by the dependence of tumor cells on a functional TP53 to maintain genomic stability when ATR is inhibited [15], as well as the importance of the ATR-CHK1 axis for G2/M checkpoint control in response to DNA damage when TP53 is mutated, thus harnessing the synthetic lethal relationship between ATR and TP53 in TP53-mutant tumors. Since mutations of the TP53 gene are present in approximately 50% of NSCLC [16], ATR inhibition represents a potential therapeutic combination strategy for DNA-damaging chemotherapy in pretreated NSCLC. In addition, recent studies have suggested that molecular alterations in other genes, such as ATM, switch/sucrose non-fermentable related, matrix associated, actin dependent regulator of chromatin, subfamily A, member 4 (SMARCA4) and AT-rich interaction domain (ARID1A), could be potential predictive biomarkers of ATR inhibition by exploiting mechanisms of synthetic lethality related to the DDR or to replication fork stability [17][18][19]. However, when leveraging the synthetic lethal relationship between ATR and ATM, it may be important to consider the emerging role of ATM in promoting tumor cell ferroptosis [20]. SMARCA4 is frequently mutated in NSCLC and is involved in the activation of replication stress responses, while ARID1A mutations increase tumor cells reliance on ATR-mediated checkpoint activity. Furthermore, ARID1A-mutant tumor cells may be more susceptible to oxidative stress due to low levels of antioxidant factors such as glutathione [21].
This phase 1 study was separated into six parts (A, B, B2, C1, C2, and C3). In the dose escalation part of this study with berzosertib and gemcitabine (part A), the most common treatment-emergent adverse events (TEAE) of any grade included fatigue, nausea, anemia, and increases in alanine aminotransferase (ALT), and the most common grade ≥ 3 TEAEs were neutropenia, increases in ALT and fatigue. These TEAEs were consistent with patient populations treated with gemcitabine [22].
The main purpose of this expansion cohort study (part C1) was to evaluate the safety, tolerability, pharmacokinetics (PK), and preliminary efficacy of berzosertib combined with gemcitabine in patients with advanced NSCLC, with or without TP53 mutations. An exploratory analysis of potential response biomarkers was also conducted (Clin-icalTrials.gov, identifier: NCT02157792).

Study design
This trial was part of a multicenter, open-label, non-randomized, phase 1 study separated into six parts (A, B, B2, C1, C2, and C3). The initial dose escalation phase of the study (parts A and B) established the recommended phase 2 dose (RP2D) of berzosertib when combined with chemotherapeutic agents, including gemcitabine and cisplatin [22,23]. These doses were further evaluated in the expansion phase of the study in patients with NSCLC (part C1), triple-negative breast cancer (part C2), and small-cell lung cancer (part C3). The focus of this manuscript is part C1; the other parts have been or will be reported separately.
Part C1 was a single-arm expansion cohort study of berzosertib combined with gemcitabine in patients with advanced NSCLC, with or without TP53 mutations. This part of the study was conducted across three sites in the UK and eight sites in the USA between December 2015 and March 2020. The study was conducted in accordance with the ethical principles of the International Council for Harmonization Guidelines for Good Clinical Practice and the Declaration of Helsinki, as well as with applicable local regulations.

Patients
The plan was to enroll approximately 40 patients, including at least 20 patients with TP53 mutations prospectively determined from archival tumor biopsies.
Eligible patients were adults ≥ 18 years of age with advanced (metastatic or locally advanced unresectable tumors and not eligible for definitive treatment), histologically confirmed NSCLC, with available archival tumor biopsies, who were intolerant to standard approved targeted therapies, and measurable disease defined by RECIST v1.1 [24]. Patients who had received more than two lines of cytotoxic chemotherapy in the advanced setting were excluded. Patients who had received treatment with gemcitabine within 6 months were also excluded. Full inclusion and exclusion criteria are shown in the Supplementary Information.

Treatments
Following screening and baseline evaluations, patients received berzosertib IV (210 mg/m 2 ; days 2 and 9) approximately 24 h after gemcitabine (1000 mg/m 2 ; days 1 and 8) in 21-day cycles, which was the RP2D established in part A of this study [22]. The timing of berzosertib relative to gemcitabine was based on the synergy demonstrated when berzosertib was administered 12-24 h after gemcitabine in preclinical models [25]. Patients received treatment until progressive disease (PD) or unacceptable toxicity.

Objectives
The primary objectives of this study were to evaluate the safety, tolerability, and the objective response rate (ORR) of berzosertib when combined with gemcitabine in patients with advanced NSCLC, with and without TP53 mutations. The secondary objectives were to evaluate the preliminary efficacy and PK of berzosertib combined with gemcitabine. The evaluation of potential biomarkers associated with the efficacy of berzosertib in combination with gemcitabine was an exploratory objective.

Assessments and endpoints
The safety profile was continuously monitored clinically and with standard laboratory parameters. TEAEs were coded according to the Medical Dictionary for Regulatory Activities (MedDRA) v21.0 [26] and graded according to the National Cancer Institute (NCI) Common Terminology Criteria for Adverse Events (CTCAE) v4.0 [27].
To evaluate the efficacy of berzosertib in combination with gemcitabine, tumor assessments were performed every two cycles for the first 12 cycles, then every two to three cycles, and 5 (±1) weeks after completion of therapy. Tumor response assessments were classified according to RECIST v.1.1 [24]. The ORR (primary efficacy endpoint) was defined as the proportion of participants who achieved a best overall response (BOR) of partial response (PR) or complete response (CR) (summarized as objective response [OR]), where both CR and PR were confirmed by repeat assessments performed no < 4 weeks after the criteria for response were first met. The ORR was calculated with the two-sided 90% confidence interval (CI) using the Clopper-Pearson method [28].
The efficacy of berzosertib in combination with gemcitabine was further explored through the assessment of progression-free survival (PFS), duration of response (DOR), overall survival (OS), and disease control.
Blood samples for PK analysis of berzosertib in plasma were collected pre-dose, mid-infusion, and 0, 0.5, 1, 2, 3, and optionally 7 h after the end of the 1-hour berzosertib infusion on cycle 1 day 2. Bioanalysis in plasma was performed using validated liquid chromatography-tandem mass spectrometry (LC-MS/MS) methods [29].
Archival tumor biopsies were collected to assess TP53 status and other genetic alterations by DNA next-generation sequencing (NGS) with FoundationOne® CDx NGS assay (Foundation Medicine, Cambridge, Massachusetts, US) [30]. A post-hoc exploratory analysis was conducted to investigate the potential association between specific genetic tumor alterations and treatment outcomes. The loss of heterozygosity (LOH), tumor mutational burden (TMB) and microsatellite instability (MSI) results were discretized according to the criteria established by Foundation Medicine [31]. LOH levels were reported as a percentage of the affected genome and were discretized to either high LOH (LOH score ≥ 16) or low LOH (LOH score < 16). TMB levels were classified as high, if somatic mutations per megabase (MB) were ≥ 20; intermediate, if somatic mutations per MB were ≥ 6 and < 20; and low, if somatic mutations per MB were < 6.

Statistical analysis
Based on historical response rates of 10% for single-agent gemcitabine in second-line NSCLC [5], and an enrollment of 30 patients, six responders would result in an exact one-sided 90% CI for ORR of 9.1-35.7%. The probability to observe at least six responders was calculated under the assumption of different response rates. In case of a true response rate of 30%, the likelihood of at least six responders was 84%.
The modified full analysis set included all patients who had a baseline tumor assessment with a measurable target lesion and at least one dose of the study drug. The safety analysis set included all patients who received at least one dose of study drug. The PK analysis set included all patients who received at least one dose of study drug and who provided at least one measurable post-dose concentration.
Summary statistics were provided for berzosertib concentrations by group and time and for berzosertib PK parameters. PK parameters were calculated using standard non-compartmental methods and the actual administered dose. Computation of PK parameters was performed using Phoenix® WinNonlin® Version 8.0 (Certara, L.P., Princeton, New Jersey, USA).

Patient demographics and disposition
Baseline and disease history characteristics of all 38 patients enrolled are presented in Table 1. For those patients whose baseline genotype was determined, TP53 mutations were found in 60.5% of tumors. All patients except for one (37,97.4%) received at least one dose of berzosertib and 38 (100.0%) patients received at least one dose of gemcitabine.

Efficacy
The median treatment duration for berzosertib in combination with gemcitabine was 14.0 weeks (2.0-63.0 weeks). There were four confirmed partial responders (10.5%), two of which had a particularly notable response (Table 3, Fig. 1).
Amongst the responders, it is worth noting that a 67-year-old female with epidermal growth factor receptor (EGFR) wild-type NSCLC (adenocarcinoma), with evidence of lymph node and lung metastases, achieved a confirmed PR lasting 13.2 months (57.6% maximum tumor shrinkage). This patient was heavily pretreated with several different anticancer regimens (carboplatin + pemetrexed + bevacizumab, followed by pemetrexed + bevacizumab as maintenance, nivolumab, and erlotinib).
A 75-year-old male patient with anaplastic large-cell lymphoma kinase wildtype, EGFR wild-type, programmed death-ligand 1 negative NSCLC (adenocarcinoma), with evidence of lymph node metastases, with progression through two aggressive regimens (carboplatin + pemetrexed, docetaxel + nintedanib followed by nintedanib maintenance), achieved a confirmed, durable PR lasting 6.0 months.

Exploratory biomarker analyses
Archival tumor biopsies from the 38 patients were analyzed by DNA NGS, of which nine samples did not pass the laboratory quality check process. Overall, 29 samples were included in the data analysis. The subgroup analysis did not demonstrate a clear association between clinical outcome (ORR and PFS) and any alterations in 324 genes explored, including TP53, and other potential predictive biomarkers of sensitivity to ATR inhibition including ATM, ARID1A (Supplementary Table 1) and SMARCA4 (Table 4).

Pharmacokinetics
The berzosertib 210 mg/m 2 (IV) dose administered in this study was within the dose range previously shown to exhibit dose-dependent berzosertib PK as monotherapy, or in combination with either carboplatin or gemcitabine [11,22]. The observed berzosertib concentration data in this expansion cohort were consistent with those reported previously at the same dose level [22]. Gemcitabine demonstrated no apparent effect on berzosertib pharmacokinetics (Fig. 2).
The geometric mean (percentage coefficient of variation [%CV]) maximum observed concentration (C max ) of berzosertib was 882 ng/mL (55.2%), which was similar to the previously reported C max of 899 ng/ mL in part A of this study [22]. A population PK model was developed based on pooled data from two phase 1 studies, including this expansion cohort [32]. The model confirmed that gemcitabine had no apparent effect on berzosertib PK, and that berzosertib PK in patients with NSCLC was comparable to patients with other advanced solid tumors.

Discussion
In this phase 1b expansion cohort study, the combination of the ATR inhibitor berzosertib with gemcitabine, according to the dosing regimen previously determined in the dose escalation portion of the trial (berzosertib 210 mg/m 2 [days 2 and 9] and gemcitabine 1000 mg/m 2 [days 1 and 8] every 3 weeks) [22], was well tolerated in patients with pretreated advanced NSCLC. The safety profile was consistent with that of the individual agents [11,23]. However, the observed clinical efficacy (ORR of 10.5% and median treatment duration of 14.0 weeks) suggests limited benefit of combining an ATR inhibitor with gemcitabine in this unselected population of patients with advanced NSCLC, with and without TP53 mutations, when compared with historical controls of gemcitabine monotherapy [5]. The ORR was only marginally higher in the subgroup of patients with TP53 mutations, because there was no responder among the few cases who did not have TP53 alterations (the study was designed to enrich for tumors with TP53 mutations). This clinical finding suggests that TP53 mutations alone are insufficient to enhance the efficacy of the berzosertib-gemcitabine combination in patients with advanced NSCLC, despite the fact that preclinical experiments have highlighted TP53 mutations as an efficacy surrogate for treatment with ATR inhibitors [14].
The exploratory biomarker subgroup analysis also demonstrated no clear association between treatment outcome (ORR and PFS) and gene alterations, including ATM, ARID1A and SMARCA4, which have previously been associated with sensitivity to ATR inhibition [17][18][19]. There was also no association observed between treatment response and alterations in other cell cycle genes such as CCNE1, MYC, or RB, whose dysregulation is associated with DNA replication stress [33,34].
However, since only single digit cases carrying genetic alterations in each of these genes were identified, the relationship between treatment response and these individual genomic alterations could not be determined. Additionally, there were limitations to our analyses, including the lack of genomic data that would have enabled the evaluation of biomarker zygosity status of tumor suppressor genes on treatment outcomes. Another limitation was the lack of confirmation of the observed ATM, SMARCA4, ARID1A or other tumor suppressor loss at the protein level. Although ATM immunohistochemistry was planned, the available tumor samples were largely exhausted after DNA NGS. As the clinical development of berzosertib continues, further investigations are required to identify genomic alterations that confer susceptibility to ATR inhibition. Identifying such alterations, with confirmation at the protein level, may ultimately help define patient populations most likely to benefit from the addition of berzosertib to DNA damage-inducing chemotherapies or other anticancer therapies.
We observed a trend towards increased response rate in patients with high TMB (33.0%) and LOH scores (30.0%), versus those patients with low TMB and LOH scores, respectively. TMB and LOH are markers of genetic instability and homologous recombination deficiency [35,36]. High TMB is also indicative of DNA DSB repair deficiency [37]. In fact, both TMB and LOH are emerging as predictive biomarkers to poly (ADPribose) polymerase and immune checkpoint inhibitors [36,[38][39][40], which could well synergize with increased DNA damage resulting from the combination of berzosertib and gemcitabine. However, the associations were not statistically significant, likely due to the small sample size. Nevertheless, given the biologic rationale for an association Fig. 1. Best percentage change in tumor size from baseline (modified full analysis set) with genetic profiles. Only patients with a baseline scan, at least one postbaseline assessment, and at least one response assessment are included in Fig. 1 (n = 34). Only patients with biomarker status determined by FoundationOne® CDx next-generation sequencing were reported. The dashed line at 20% represents PD whereas the dashed line at -30% represents PR. ARID1A, AT-rich interaction domain 1A; ATM, ataxia telangiectasia mutated; LOH, loss of heterozygosity; NE, not evaluable; PD, progressive disease; PR, partial response; SMARCA4, switch/ sucrose non-fermentable related, matrix associated, actin dependent regulator of chromatin, subfamily A, member 4; SD, stable disease; TMB, tumor mutational burden; TP53, tumor protein p53. between TMB and LOH and a higher sensitivity to DDR inhibitors, further investigations in this direction are warranted.

Conclusions
The combination of berzosertib and gemcitabine in patients with advanced, pre-treated NSCLC was well tolerated, but given the observed clinical efficacy, future clinical trials may best be undertaken in a genomically selected patient population. In other malignancies, such as platinum-resistant ovarian cancer, the combination of berzosertib and gemcitabine has shown an encouraging efficacy signal, serving as a reminder of the molecular heterogeneity and notable clinical differences between disease entities [12].

Role of the funding source
The trial was sponsored by Merck Healthcare KGaA, Darmstadt, Germany (CrossRef Funder ID: https://doi.org/10.13039/100009945) and Vertex Pharmaceuticals Incorporated., Boston, MA, USA. Both sponsors were involved in the development of the study design and the collection of data. Merck Healthcare KGaA, Darmstadt, Germany, provided financial support for the preparation of this article, and provided technical assistance for the analysis and interpretation of data. The decision to submit the article for publication was made by the authors together with Merck Healthcare KGaA, Darmstadt, Germany.   Ivan Diaz-Padilla: is a former employee of Ares Trading SA, Eysins, Switzerland, an affiliate of Merck KGaA, Darmstadt, Germany; is a current employee of GlaxoSmithKline, Zug, Switzerland; and received medical writing support for developing this manuscript from Merck Healthcare KGaA, Darmstadt Germany. Geoffrey I Shapiro: reports grants/contracts from Eli Lilly, Merck & Co., Inc., Sierra Oncology and Pfizer; holds a patent entitled, 'Dosage regimen for sapacitabine and seliciclib', also issued to Cyclacel Pharmaceuticals, and a pending patent, entitled, 'Compositions and Methods for Predicting Response and Resistance to CDK4/6 Inhibition'; received medical writing support from Pfizer; and received medical writing support for developing this manuscript from Merck Healthcare KGaA, Darmstadt Germany.