Non-linear IV pharmacokinetics of the ATR inhibitor berzosertib (M6620) in mice

Abstract

Background:

The Ataxia Telangiectasia and Rad3-related (ATR) protein complex is an apical initiator of DNA damage response pathways. Several ATR inhibitors (ATRi) are in clinical development including berzosertib (formerly M6620, VX-970). Although clinical studies have examined plasma pharmacokinetics (PK) in humans, little is known regarding dose/exposure relationships and tissue distribution. To understand these concepts, we extensively characterized the PK of berzosertib in mouse plasma and tissues.

Methods:

A highly sensitive LC-MS/MS method was utilized to quantitate berzosertib in plasma and tissues. Dose proportionality was assessed in female BALB/c mice following single IV doses (2, 6, 20 or 60 mg/kg). A more extensive PK study was conducted in tumor-bearing mice following a single IV dose of 20 mg/kg to evaluate distribution to tissues. PK parameters were calculated by non-compartmental analysis (NCA). A compartmental model was developed to describe the PK behavior of berzosertib. Plasma protein binding was determined in vitro.

Results:

Increased doses of berzosertib were associated with less than proportional increases in early plasma concentrations and greater than proportional increase in tissue exposure, attributable to saturation of plasma protein binding. Berzosertib extensively distributed into bone marrow, tumor, thymus, and lymph nodes, however; brain and spinal cord exposure was less than plasma.

Conclusion:

The nonlinear PK of berzosertib displayed here can be attributed to saturation of plasma protein binding and occurred at concentrations close to those observed in clinical trials. Our results will help to understand preclinical pharmacodynamic and toxicity data and to inform optimal dosing and deployment of berzosertib.

1. INTRODUCTION

Cancer therapy compromises chemotherapeutic agents and ionizing radiation (IR) that induce DNA damage and cell death in both healthy and malignant tissue. The DNA damage response (DDR) is a signaling system that integrates the cell cycle with DNA repair to safeguard genome stability and limit cell and tissue damage [1,2]. An apical initiator of DDR pathways is Ataxia Telangiectasia and Rad3-related (ATR), a serine/threonine protein kinase that is recruited to and activated at a variety of DNA damage sites including RPA-coated singe-stranded DNA (ssDNA) at stalled and collapsed replication forks [2,3]. ATR-dependent phosphorylation of CHK1 activates this downstream kinase and extends DDR processes through phosphorylation of CDC25 phosphatases and inactivation of cyclin-dependent kinases, resulting in cell cycle arrest [4–6]. There is an increasing interest in targeting ATR with small molecule inhibitors to potentiate the anticancer effects of cytotoxic chemotherapies and IR.

Small molecule ATR inhibitors (ATRi) show promising antitumor activity in multiple preclinical models when combined with DNA-damaging agents, such as IR and immune checkpoint inhibitors (ICI) [7–9]. For example, it has been previously shown that ATRi ceralasertib potentiates tumor-specific CD8+ T-cell responses following radiation in mice [8]. However, questions regarding the ability of ATRi to induce cancer cell death, without exacerbating damage to otherwise healthy tissue remain unanswered. This is particularly relevant when considering the narrow efficacy and toxicity thresholds associated with chemotherapy and RT. Therefore, accurately capturing and defining ATRi drug exposure is critical to elucidating its ability to shift therapeutic indices, ultimately impacting the agent’s optimal use.

Several small molecule ATRi are in various stages of clinical development, including the intravenously (IV) administered berzosertib (M6620, VX-970) which is currently being investigated in 17 clinical trials (Clinicaltrials.gov accessed 02/08/2024). Preclinical and clinical investigations of berzosertib have shown it to exert antitumor activity as both a monotherapy and in combination with DNA-damaging agents, such as cisplatin and IR, especially in p53-deficient cell lines [10–12]. Although clinical studies have reported plasma pharmacokinetics (PK) in humans, little is known regarding dose-exposure relationships and tissue distribution which may determine its therapeutic window. Furthermore, it is unknown if berzosertib will induce immunological responses following IR in mice to a similar extent which was seen with ceralasertib [7,8]. Thorough analysis of tissue distribution may identify doses and exposures associated with efficacy and toxicity. To improve cancer therapies and support clinical development of ATRi, several berzosertib PK studies were completed in mice to comprehensively determine distribution to tumor and other tissues. Results presented here will provide future investigators with a PK framework that will help interpret differences in efficacy and toxicity with other ATRi in development [13,14].

2. MATERIALS AND METHODS

2.1. Chemicals and Reagents

Berzosertib was purchased from AdooQ Bioscience (Irvine, CA) and [D₇]-berzosertib (Suppl.Figure 1) was custom synthesized and purchased from ALSACHIM (Illkirch-Graffenstaden, France). Formic acid and HPLC grade water and methanol were obtained through Fisher Scientific (Fairlawn, NJ). DMSO was obtained from Sigma-Aldrich (St. Louis, MO). For use in vehicle formulation, Captisol^® (Sulfobutylether-β-Cyclodextrin) was purchased from Selleck Chemicals (Houston, TX) and acetic acid was purchased from Fisher Scientific (Fairlawn, NJ).

2.2. Mice

Specific pathogen-free female BALB/c mice (5–7 weeks of age) were purchased from Charles River (Wilmington, MA). The mice were handled in accordance with Guide for the Care and Use of Laboratory Animals (National Research Council, 2011) standards and on a protocol approved by the University of Pittsburgh IACUC. Mice were allowed to acclimate to the University of Pittsburgh Animal Facility for at least 1 week prior to study initiation. To minimize exogenous infection, mice were maintained in microisolator cages where ventilation and airflow were set to 12 changes/h. Rooms were kept on automatic 12-h light/dark cycles and temperature was regulated at 72±4 °F. The mice received Prolab ISOPRO RMH 3000, Irradiated Lab Diet (PMI Nutrition International, Brentwood, MO) and water ad libitium. Prior to the study, mice were stratified primarily by body weight and secondarily by tumor size (when applicable) to ensure appropriate distribution at each time point. Throughout all studies, mice were routinely weighed and monitored for changes in health.

2.3. Dose Linearity Study

The dose linearity study was performed within 2 days using a single batch of mice (N=3 per time point) dosed with 2.0, 6.0, 20 or 60 mg/kg berzosertib IV via a ~30 s tail vein bolus. The vehicle was 12% captisol adjusted to a pH of 4 with acetic acid. Sample collection time points were 5, 15, 60, 120, 360 and 1,440 min. Vehicle control was also included at 5 min. Mice were euthanized by CO₂ inhalation followed by thoracotomy. The following tissues were collected at each time point: blood, liver, kidney, lung, skeletal muscle, and brain. Blood was collected by cardiac puncture using EDTA anticoagulated syringes, transferred to microcentrifuge tubes, and centrifuged at 12,000 × g for 4 min to separate plasma and red blood cells (RBCs). Pooled urine and feces were collected over ice from mice housed in metabolic cages and euthanized at 24 h.

2.4. Extensive Tissue Distribution Study

Mice were injected subcutaneously on the right flank with approximately 1×10⁶ CT26 (murine colorectal carcinoma) cells obtained from ATCC (Manassas, VA). Cells were cultured in RPMI-1640 medium with L-glutamine (BioWhittaker Inc., Walkersville, MD), containing 10% heat-inactivated fetal bovine serum and 100 units of penicillin/mL and 100 μg/mL of streptomycin (Biofluids, BioSource, Rockville, MD) in an incubator with 95% air, 5% CO₂, and 95% humidity at 37 °C. Cells were checked for mycoplasma by IDEXX BioAnalytics (Westbrook, ME). Implantation and tumor growth were monitored twice weekly with a digital caliper. When tumors were approximately 200 mm³ mice were stratified into time groups (N=3) as previously described. Mice were administered 20 mg/kg via tail vein IV. Sample collection time points were 5, 15, 30, 60, 120, 240, 360, 960, and 1,440 min and a vehicle control included at 5 min. At each time point mice were euthanized and the following tissues collected: blood, liver, kidney, spleen, lung, skeletal muscle, brain, heart, fat, tumor, small intestine (flushed with PBS), esophagus, spinal cord, thymus, draining lymph node, non-draining lymph node, and bone marrow. Bone marrow was obtained by flushing both femurs from each mouse with PBS, followed by centrifugation at 12,000 × g and removal of the supernatant. Protein analysis of the resuspended pellet was conducted using the Bio-Rad protein assay following the manufacturer’s instructions with bovine serum albumin as the standard. Blood was processed to obtain plasma and urine and feces were collected in the same manner as described above.

2.5. Bioanalysis

To quantitate berzosertib in plasma and tissues, an LC-MS/MS assay was applied on a system containing an Agilent (Palo Alto, CA) 1290 Infinity II Autosampler and Binary Pump and a SCIEX (Concord, ON, Canada) 6500+ mass spectrometer. This assay was based on a previously validated method, to which modifications were made to enhance sensitivity [15]. This included lowering the lower limit of quantitation (LLOQ) 6-fold (from 3 ng/mL to 0.5 ng/mL), adjusting the QC ranges to reflect the increase in sensitivity and alterations to sample preparation to accommodate tissue homogenization.

Conditions for the mass spectrometer were set following infusion of neat standards and eventually optimized. Modifications to the previously published method included: 20 L/h curtain gas, 5,000 IS, 300 °C TEM, 70 L/h GS1 and GS2, medium CAD, 10 V EP, 20 V CE, 10 V CXP. MRM channels were m/z 464.1>433.0 for berzosertib and m/z 471.1>439.9 for the internal standard, [D₇]-berzosertib, with dwell times of 0.1 s for both analytes. The column, mobile phase composition, and gradient were unchanged from the previous method [15].

Duplicate standard curves with calibrators consisting of 0.5, 1, 5, 10, 50 100, 500, 1,000 and 5,000 ng/mL, in addition to a blank plasma control, were prepared fresh on days of analysis using control BALB/c mouse plasma (Innovative Research, Novi, MI). QCs were prepared in bulk at concentrations of 1.5 (QCL), 150 (QCM), and 4,000 ng/mL (QCH), aliquoted, and stored at −80 °C. Sample preparation was carried out by adding 10 μL of 50 ng/mL [D₇]-berzosertib to each 50 μL sample followed by the addition of 200 μL of acetonitrile for protein precipitation. Samples were then vortexed at maximum speed for 1 minute and centrifuged for 5 min at 13,500 × g. 100 μL of the resulting supernatant was then transferred to HPLC vials followed by 150 μL of HPLC-grade water, which was then briefly vortexed. The injection volume was 5 μL with a total run time of 4.0 min. Retention times were 1.12 min and 1.11 min for berzosertib and [D₇]-berzosertib, respectively. Curve fitting was accomplished through linear regression with 1/y² weighting. Concentrations were determined from calibration curves prepared in control mouse plasma.

For tissue and tumor analysis, samples were homogenized with 3 parts PBS (v/g) and further diluted with control mouse plasma as needed. To validate the assay, a duplicate standard curve with QCs (N=3, per level) was analyzed to determine analytical accuracy and precision. Assay performance in matrices of tissue homogenates was determined by spiking control tumor, liver, kidney, spleen, heart, fat, muscle, brain, and small intestine samples in triplicate. Recovery was calculated using a calibration curve constructed in mouse plasma.

2.6. Pharmacokinetic Analyses

2.6.1. Noncompartmental Analysis

PK parameters of berzosertib were estimated by non-compartmental analysis (NCA) using Phoenix WinNonlin^® software (version 8.1, Certara, Inc., Princeton, NJ, USA) and by means of the Bailer method for AUC and standard error of the mean (SEM) determination in plasma, RBC, tumor and tissues [16]. Tissue partition coefficients (P_Tissue) were calculated using Eq. (1) and tested by ANOVA with Dunnett’s multiple comparisons post-hoc test, as was dose normalized AUC where the 2 mg/kg cohort was used as the control group. Dose normalized C_max was analyzed by Jonckheere-Terpstra trend test.

To statistically evaluate dose linearity we used Eq. (2), a power model integrating dose with exposure which, following natural logarithmic (Ln) transformation (Eq. (3)), was used to assess proportional relationships between Ln(y), exposure, and Ln(dose) [17]. Ln-transformed tissue AUC_0-∞ was plotted versus Ln-transformed dose and a linear regression was used to calculate a slope that was subsequently compared to a slope of β=1, a value that would indicate linearity between dose and exposure.

For analysis of excreta, urine was diluted at least 1:10 in control mouse plasma and feces were homogenized in 1:3 in PBS (g/v). To obtain the total amount of drug in excreta, the resulting concentrations were multiplied by the volume or weight of urine or feces, respectively, then divided by 3 to account for the number of contributing mice in each cage. The amounts were subsequently divided by the amount of administered drug to yield a percent excreted in urine and feces (Eq. (4)).

2.7. Plasma Protein Binding

Plasma protein binding of berzosertib was assessed with rapid equilibrium dialysis (RED) devices (Thermo Fisher Scientific, Waltham, MA). Freshly spiked BALB/c mouse plasma (300 μL) and PBS (500 μL) were added to the sample and buffer chambers, respectively. Samples were incubated at 37 °C for 24 h in triplicate at 300, 1,000, 3,000, 10,000, 30,000 and 100,000 ng/mL with <0.1% final organic solvent content.

After determining the fraction unbound in plasma (f_u,p), the concentration of unbound and bound drug in each sample was calculated. Assuming berzosertib binds to albumin, the most abundant plasma protein which has an approximate concentration of 406 μM in BALB/C mice, we calculated the concentration of unbound protein using Eq. (5) [18]. The dissociation constant (K_d) was then calculated using Eq. (6).

2.8. Compartmental Model

A unified compartmental PK model was developed in ADAPT5 to fit the data generated from the dose linearity study [19]. Naïve-pooled concentration-time data from each cohort were used to estimate PK parameters shared between each treatment group, initially with a linear model. Next, in vitro protein binding data was used to add a component within the central compartment, allowing for saturable protein binding. The f_u,p was calculated using Eq. (7). K_d and the estimation of total plasma protein (P₀) was made under the assumption of albumin binding to berzosertib. Total drug concentration (R₀) was informed using our mouse data and total plasma albumin concentration (P₀) was set to a fixed literature value [18]. The model was structured such that unbound drug concentration drove egress into the peripheral compartment. Differential equations used in each model can be seen in Eqs. (8)–(11). Model performance was assessed by visual inspection of standardized residual concentration-time plots, coefficient of variation (CV%), goodness of fit (R²), and Akaike Information Criterion (AIC).

3. RESULTS

A representative chromatogram of a berzosertib LLOQ sample from the cross-validated LC-MS/MS method can be seen in Suppl.Figure 2. The accuracy and precision of a duplicate standard curve and QCs met the acceptance criteria for FDA bioanalytical method validation with accuracies of calibrators and QCs ranging between 86.9 to 111% with precisions ≤8.4% (Suppl.Table 1) [20]. The recovery of individual spiked tissue homogenates quantitated with a plasma calibration curve were: tumor (93.0%), liver (97.3%), kidney (105%), spleen (99.4%), heart (94.8%), fat (111%), muscle (99.1%), brain (95.9%), and small intestine (92.5%).

3.1. Dose Linearity Study

3.1.1. Noncompartmental Analysis

The plasma concentration versus time profiles of berzosertib are depicted in Figure 1A and noncompartmental PK parameters are detailed in Table 1. At all dose levels, plasma profiles appeared to be biphasic with C_max and AUC increasing with dose. Dose-normalized concentration profiles (Figure 1B) appeared similar up to 120 min. The 2 and 6 mg/kg dose normalized concentration profiles were largely overlapping, while 20 mg/kg fell lower and 60 mg/kg was the lowest. This rank-order was also observed in dose normalized C_max and dose normalized AUC_0–360 values (Figure 1C, D). Observed terminal half-lives (t_1/2) increased with dose and ranged from 153 to 629 min, although this is in part attributable to different times of last observation above the LLOQ. Dose escalation was associated with a decrease in clearance (CL) when calculated using AUC_0-∞ (38.5 – 28.3 mL/min/kg) and an increase in CL when calculated using AUC_0–360 (43.6 – 64.6 mL/min/kg). A more than 3-fold increase in volume of distribution (V_d) (8.49 – 25.6 L/kg) was observed between 2 and 60 mg/kg, with the biggest change occurring between 20 and 60 mg/kg. Tissue concentration versus time profiles, shown in Figure 2A–D, appeared to mirror the biphasic plasma profile with T_max occurring at 5 min for the majority of tissues, suggesting perfusion-rate limited distribution. All tissue samples provided observable concentrations up to 1,440 min, with only skeletal muscle and brain falling below the LLOQ at lower dose levels. Dose-normalized tissue concentration versus time profiles were similar up to ~120 min and started to diverge at later time points resulting in shallower terminal slopes (longer half-lives) as dose increased (Suppl.Figure 3A–E). AUC increased with dose in all tissues (Suppl.Table 2) and, when normalized to dose, exhibited non-linear exposure in the opposite direction of that seen in plasma (Suppl.Figure 4A–F) with the greatest differences occurring within the 20 and 60 mg/kg treated groups. Tissue partition coefficients (Figure 3A–D) increased with dose with statistically significant differences between the 60 mg/kg and the 2 mg/kg group in all tissues. Brain also exhibited significantly increased partitioning at 20 mg/kg. Tissue distribution higher than in plasma (partition coefficient > 1) was observed in kidney, lung and skeletal muscle across all dose levels, while for RBC and brain, this only occurred starting at 20 and 60 mg/kg, respectively (Suppl.Table 3).

Figure 1.

IV PK and NCA of berzosertib from the dose linearity study. A) Mean plasma concentration versus time profiles for 2 mg/kg (○), 6 mg/kg (□), 20 mg/kg (△), 60 mg/kg (◇). B) Mean plasma concentrations normalized by administered dose versus time profiles C) Dose-normalized plasma C_max. D) Dose-normalized plasma AUC_0–360. Error bars represent ± SD for concentrations and ± SEM for AUC. AUC_0–360 calculated using the Bailer method [16]. *P < 0.05 by Jonckheere-Terpstra trend test for C_max/Dose and **P < 0.01 one-way ANOVA with Dunnett’s multiple-comparison test for AUC_0–360/Dose.

Table 1.

Noncompartmental berzosertib PK parameters.

Route	IV	IV	IV	IV	IV-exta
Dose (mg/kg)	2.0	6.0	20	60	20
Cmax (μg/mL)	1.48 (0.706)	3.47 (1.47)	10.2 (3.67)	21.2 (4.46)	13.3 (4.91)
Cmax/dose	0.738	0.578	0.512	0.353	666
Tmax (min)	5	5	5	5	5
Tlast (min)	360	360	1440	1440	1440
AUC0–360 (μg/mL•min)	45.9	126	424	929	591
AUC0–360/dose	23.0	21.1	21.2	15.5	29.6
AUC0-t (μg/mL•min)	45.9	126	605	1,762	854
AUC0-t/dose	23.0	21.1	30.3	29.4	42.7
AUC0-∞ (μg/mL•min)	52.0	147	614	2124	906
AUC0-∞ /dose	26.0	24.6	30.7	35.4	45.3
AUCextrap(%)	11.7	14.2	1.39	17.0	5.75
CLb (mL/min/kg)	43.6	47.6	47.2	64.6	33.8
CLc (mL/min/kg)	38.5	40.7	32.6	28.3	22.1
Vd (L/kg)d	8.49	9.63	11.9	25.6	11.7
Half-life (min)d	153	164	253	629	367
0–24 h urine (% dose)	40.6	49.2	55.8	28.0	48.4
0–24 h feces (% dose)	1.78	3.41	1.48	2.02	4.13

Values represent mean values (SD for C_max)

^a.Plasma PK parameters from extensive tissue distribution study

^b.Calculated using AUC_0–360

^c.Calculated using AUC_0-∞

^d.For doses 2 and 6 mg/kg, these parameters were calculated based on data out to 360 min, while for the 20 and 60 mg/kg dose, these parameters were calculated based on data out to 1440 min

Figure 2.

Mean plasma concentration versus time profiles from the dose linearity study for A) Kidney B) Lung C) Skeletal Muscle D) Brain. 2 mg/kg (○), 6 mg/kg (□), 20 mg/kg (△), 60 mg/kg (◇). Error bars represent ± SD.

Figure 3.

Tissue partition coefficients from the dose linearity study. A) Kidney B) Lung C) Skeletal Muscle D) Brain. Partition coefficients were based on AUC_0–360 calculated with the Bailer method [16]. Error bars represent ± SEM. *P < 0.05, ***P < 0.001, ****P < 0.0001 by one-way ANOVA with Dunnett’s multiple-comparison test.

Plots of Ln(AUC_0-∞) versus Ln(dose) are shown in Figure 4A–D. Following linear regression, using the power model in Eq (3), all tissue groups displayed linear slopes >1, suggesting greater than proportional increases in exposure with increasing dose. The greatest deviation from 1 was detected in skeletal muscle, with a slope of 1.43 which was also the only tissue where statistical significance was observed as the 95% CI (1.23–1.63) did not include the value 1.

Figure 4.

Ln-transformation of tissue AUC_0-∞ versus Ln-transformation of dose. A) Kidney: β=1.35 (95% CI: 0.960 – 1.73) B) Lung: β=1.30 (95% CI: 0.804 – 1.79) C) Skeletal Muscle: β=1.43 (95% CI: 1.23 – 1.63) D) Brain: β=1.17 (95% CI: 0.786 – 1.56). Dotted lines represent β=1.

Unchanged berzosertib in urine increased from 40.6 to 55.8% in the 2 to 20 mg/kg treated groups but decreased to 28.0% in the 60 mg/kg cohort. In feces, <4% of unchanged drug was recovered across all groups, with no noticeable trend as a function of dose (Table 1).

3.2. Extensive Tissue Distribution Study

3.2.1. Noncompartmental Analysis

Concentration-time profiles for plasma and tissues collected during the extensive tissue distribution study are shown in Suppl.Figure 5A–R. All samples provided concentrations above the 0.5 ng/mL LLOQ. The plasma concentration-time profile and PK in tumor bearing animals corresponded well with most PK of the 20 mg/kg non-tumor bearing treatment group from the dose linearity study, with notable increases in AUC and half-life that were the result of a lower clearance (Table 1). Tissue exposure and partition coefficients indicated that bone marrow, kidney, liver, spleen, lung, small intestine, tumor, non-draining lymph nodes, thymus, draining lymph node, heart, esophagus, skeletal muscle, and fat experienced greater exposure than plasma, while RBC, brain and spinal cord experienced exposure less than plasma (Table 2 and Suppl.Figure 6A–B). In the majority of tissues, drug distribution was assumed to be permeability-rate limited as observed tissue T_max was later than plasma (5 min) with the exception of RBCs, fat, heart, kidney, liver and lung (Table 2).

Table 2.

Extensive Tissue Distribution Study PK and Partition Coefficients.

Tissue	Cmaxa (μg/mL)	Tmax (min)	AUC0-tb (μg/mL•min)	AUC0-∞ (μg/mL•min)	Tissue Partition Coefficient
Plasma	13.3 (4.91)	5	854	906	1
Bone Marrow	164 (153)	60	63,581	66,660	74.5
Kidney	117 (25.4)	5	49,230	52,367	57.7
Liver	79.8 (26.7)	5	21,805	23,205	25.5
Spleen	39.0 (3.86)	120	17,571	18,375	20.6
Lung	45.2 (11.8)	5	11,889	12,959	13.9
Small Intestine	31.3 (5.80)	15	10,845	12,510	12.7
Tumor	13.1 (1.39)	60	7,413	8,378	8.68
Non-Draining Lymph Node	20.7 (8.26)	30	6,650	6,945	7.79
Thymus	13.5 (5.31)	30	5,070	5,572	5.94
Draining Lymph Node	12.9 (0.131)	60	4,818	5,222	5.70
Heart	31.6 (6.27)	5	4,869	5,189	5.64
Esophagus	12.1 (8.06)	30	2,619	2,803	3.07
Sk. Muscle	9.95 (1.41)	15	2,364	2,504	2.77
Fat	4.28 (2.55)	5	1,358	1,511	1.59
RBC	1.69 (0.957)	5	332	352	0.389
Brain	0.627 (0.079)	15	254	271	0.297
Spinal Cord	0.307 (0.102)	15	120	129	0.140

Error is presented as SD for C_max

^a.Observed parameter

^b.AUC are from 0–1440 min with corresponding tissue partition coefficient

^c.All infinity extrapolated AUC portions are < 14%

Excretory recovery was similar to observations in tumor free mice from the previous dose linearity study with 48.4% and 4.13% of unchanged berzosertib recovered in urine and feces, respectively (Table 1).

3.3. Plasma Protein Binding

The fraction of unbound berzosertib in mouse plasma (f_u,p) increased from 2.80% to 4.27% between 300 and 100,000 ng/mL (Suppl.Table 4 and Suppl.Figure 7). Assuming albumin binding and an albumin concentration of 406 μM, calculated K_d values resulted in a geometric mean of 11.9 μM (N=6 concentrations, range: 8.9 to 11.7 μM) [18].

3.4. Compartmental Model

A compartmental PK model was developed using pooled concentration-time data from all treatment groups within the dose linearity study (Suppl.Figure 8). Although both a linear and saturable 2-compartmenal model fit the data well, the saturable model resulted in a slightly better fit (Suppl.Table 5 and Suppl.Figure 9). To calculate berzosertib f_u,p (Eq. (7)), K_d and P₀ were fixed at 11.9 μM and 406 μM, respectively, based on our in vitro data and the literature [18]. Following an IV bolus, drug experienced saturable protein binding within the central compartment which allowed for only unbound drug to drive entry into to the peripheral compartment (CL_D). Following redistribution (CL_D) into the central compartment, drug was then able to be cleared (CL_T).

4. DISCUSSION

To comprehensively determine IV berzosertib exposure in plasma, tumor, and other tissues, several PK studies were conducted in mice. Results from the tissue distribution study showed that berzosertib distributed well. The majority of collected tissues experienced greater exposure than plasma, in line with volume of distribution estimates of approximately 10 L/kg and up. Dose escalation from 2 to 60 mg/kg was associated with less than proportional increases in exposure in plasma (judged by C_max and the 0–360 min time frame for which data were available for all doses) and greater than proportional in peripheral tissues. At 24 h, the 2 and 6 mg/kg cohorts were unable to provide a terminal plasma sample > LLOQ. Thus, the AUC_0-∞ is underestimated in these groups, implying decreasing drug CL between 2 and 60 mg/kg. However, dose-normalized AUC_0–360 values provided a means to compare plasma exposure across the dosing range and were consistent between 2 and 20 mg/kg (23.0 – 21.2), but decreased at 60 mg/kg (15.5). Apparent clearance, calculated with AUC_0–360, remained stable between 2 and 20 mg/kg (43.6 – 47.2 mL/min/kg) and increased at 60 mg/kg (64.6 mL/min/kg). These observations were indicative of saturable plasma protein binding which was confirmed with our in vitro experiments, at plasma concentrations observed in vivo. The dose dependent increase in free fraction at early time points shifts drug away from the plasma compartment into the tissues, and this drug reappears later in the concentration time profile upon redistribution. Unfortunately, this also means that one cannot compare at face value plasma clearance calculated based on truncated AUC values, because the distribution of actual AUC before the truncation vs AUC after the truncation changes with dose. The mechanistic explanation of non-linearity described here allows for a more robust understanding of PK and exploration of consequent relationships with observed PD endpoints, such as toxicity and efficacy.

Within the dose linearity study, berzosertib appeared to display biphasic plasma concentration-time profiles for all doses. The change in half-life across the dose range was driven by an increasing V_d, ultimately influencing an increase between the lowest (t_1/2 = 153 min) and highest dose (t_1/2 = 629 min). As the 20 and 60 mg/kg groups had observable terminal time points (T_last = 1,440 min), the increase in half-life between these two groups was most likely attributable to prolonged redistribution of drug from tissues back into the plasma. This trend was also observed with non-superimposable dose-normalized plasma concentrations and greater-than-proportional increases in dose-normalized AUC_0-∞. Drug egress from plasma into tissue is largely driven by the unbound drug concentration gradient, we therefore hoped to utilize tissue exposures to confirm the non-linearity in plasma. Tissue analyses were indeed instrumental in confirming our findings with significantly higher tissue partitioning in the 60 mg/kg treated groups compared to the 2 mg/kg treated group. The brain was unique in showing significantly higher partitioning already at 20 mg/kg. We suggest that this is because the exposure in brain, of all tissues, is the most sensitive read-out of changes in free fraction of drug. Berzosertib entry into brain is driven exclusively by diffusion across the blood-brain barrier (BBB) as the presence of tight junctions prevents drug transfer via convection (i.e., bulk flow). Thus, any change in free drug concentration directly translates to a proportional change in brain concentration. In contrast, berzosertib entry into other tissues is driven by a combination of diffusion and convection, the latter through fenestrated endothelium with drainage through the lymphatic system. Any effect on tissue concentrations by a change in free drug concentration will be diluted with the contribution of convection of total drug concentration into tissues, and therefore be harder to detect. A more speculative mechanism that could contribute to this earlier effect in brain would be saturation of efflux transporters at the BBB, as berzosertib is a known substrate for both breast cancer resistance protein (Bcrp/Abcg2) and P-glycoprotein (P-gp/Mdr1) on the BBB [22].

We were able to accurately portray the complex PK behavior of berzosertib and fit our dose linearity data to a 2-comparmental model encompassing a saturable plasma protein binding component within the central compartment, which was informed by our in vitro protein binding data. The rapid saturation of protein binding harmonized with the observed C_max which ultimately influenced the non-linearities seen at higher doses. By assuming a moderately high extraction ratio, only unbound concentrations of berzosertib were allowed to distribute to the peripheral compartment, while total drug concentrations drove clearance from the central compartment.

The amount of unchanged berzosertib collected in urine over 24 h increased with dose in the 2, 6, and 20 mg/kg treated groups (40.6 to 55.8%) but decreased to 28% at 60 mg/kg. This is most likely the result of tissue drug concentration still being relatively high at 24 h in the 60 mg/kg group. In tissues with a quantifiable 24 h sample, the AUC extrapolated beyond 24 h was < 3% in the 2, 6, and 20 mg/kg treated groups, while at 60 mg/kg, the AUC extrapolated in tissues was > 20% (range:10.2 to 24.4%). Therefore, the disproportionately high tissue partitioning at the highest dose level reduced the proportion of dose passing through the kidneys to be eliminated by 24 h, even though briefly at the very earliest time points, renal clearance will likely have been higher at the highest dose, while the free fraction was increased.

Following 20 mg/kg IV administration, berzosertib tumor T_max was delayed, likely due to poor tumor vascularization, relative to plasma and occurred at 60 min. However, tumor C_max was almost identical to plasma, with an AUC_0-∞ approximately 8-fold higher. Distribution in the thymus and lymph nodes, both central lymphoid organs, was most likely influenced by permeability rate limited kinetics as T_max also occurred after plasma. Exposure in these tissues also exceeded plasma (P_tissue > 5), highlighting an affinity of berzosertib to tissues within the immune system. Comparable concentrations were observed in both the non-draining and draining lymph nodes, similar to our observations in tumor-bearing mice receiving the orally administered ATRi, ceralasertib (AZD6738) and elimusertib (BAY-1895344) [13,14]. While berzosertib distributed well overall into peripheral tissues, investigators should consider the delayed T_max in tumors and lymph nodes in their design of future in vivo studies. Specifically, as it relates to timing the combination with DNA damaging agents, such as radiation. Bone marrow experienced a 74.5-fold higher exposure than plasma, which was the highest of all tissues collected and greater than relative plasma concentrations seen with ceralasertib (4.60-fold) and elimusertib (2.65-fold) [13,14]. Clinically, berzosertib has been associated with bone marrow toxicities including neutropenia, thrombocytopenia, leukocytopenia, lymphocytopenia, and anemia [23–27] which may be attributed to high bone marrow concentrations, similar to observations in our study.

Following BSA-based scaling, the selected IV murine doses of 2, 6, 20 and 60 mg/kg translate to doses of 6, 18, 60 and 180 mg/m² [27]. Although these fall below the recommended phase 2 dose (RP2D) of berzosertib as a monotherapy (240 mg/m²), they nicely align with doses of berzosertib (90–210 mg/m²) commonly used in clinical trials in combination with cytotoxic chemotherapy and immunotherapy [28,22–26,29–31]. Clinically, berzosertib has been reported to exhibit dose-proportionate plasma PK up to 240 mg/m², which may be attributed to lower C_max in humans compared to mice [28,24,31]. In patients receiving 60–480 mg/m² berzosertib, the C_max ranged from 0.286–4.4 μg/mL, slightly below our murine C_max values of 10–20 μg/mL where we began to observe saturated plasma protein binding. AUC values in patients that received berzosertib 90–210 mg/m² were 169–404 μg/mL•min which is lower than the AUC_0-∞ of 614–2124 μg/mL•min observed in the mice from our studies at 20 and 60 mg/kg, which would approximate the BSA-based corresponding doses [30]. The saturable murine PK was most clearly observed in our tissue data, and those data are rarely available in clinical studies. It may therefore be that in patients at the higher end of berzosertib dosing, tissue concentrations increase more than proportionally, while total plasma concentrations are only minimally affected. Clinicians should therefore closely monitor berzosertib toxicities at increasing doses or infusion speeds due to unexpectedly high tissue concentrations.

When evaluating dose proportionality, a common method is to assess natural Ln-transformations of dose and exposure (e.g., AUC) via linear regression, where dose linearity would yield a slope of 1. In our study, all tissues collected displayed slopes larger than 1, suggesting greater than dose proportional increases in exposure. Skeletal muscle was the only tissue where the 95% confidence interval excluded the value of 1, meeting statistical significance. An alternative method for assessing dose proportionality is to statistically compare tissue coefficients across the dose range. Partition coefficients for all tissues collected in the dose linearity study were largely uniform across the two lowest doses, but greatly increased thereafter with statistically significant increases between the lowest and highest dose. Although both methods suggest a nonlinear relationship between dose and exposure, the discrepancy between statistical significance in tissues beyond skeletal muscle can be attributed to the incorporation of standard error (SEM) when comparing tissue partition coefficients, information that is not utilized in the power model approach which utilizes just a single value per dose. The confirmation of nonlinear kinetics is also consistent with non-superimposable plasma and tissue concentration-time profiles. In vitro measurement of berzosertib plasma protein binding showed that the greatest increase in f_u,p occurred between 3 and 10 μg/mL, just below the observed plasma C_max in the 20 mg/kg treated group (10.2 μg/mL at 5 min post-dose). As plasma C_max more than doubles in the 60 mg/kg treated group (21.2 μg/mL), saturable plasma protein binding explains the nonlinearities observed at the two highest doses. It should be noted that our plasma C_max merely represents the concentration documented at the earliest sample time of 5 min, while immediately after dosing the actual concentrations in the mice are likely to be even higher.

CONCLUSION

IV berzosertib displayed nonlinear PK in mice at concentrations close to those observed in clinical trials. Increased doses of berzosertib were associated with less than proportional increases in early plasma concentrations and greater than proportional increase in tissue exposure, attributable to saturation of plasma protein binding. Our results will help to better understand preclinical pharmacodynamic and toxicity data and will help to inform optimal dosing and clinical deployment of berzosertib.

Abbreviations

ATR

Ataxia telangiectasia and Rad3-related

ATRi

Ataxia telangiectasia and Rad3-related inhibitor

DDR

DNA damage response

LC-MS/MS

Liquid chromatography-tandem mass spectrometry

LLOQ

Lower limit of quantitation

PK

Pharmacokinetics

NCA

Noncompartmental Analysis

C_max

Maximum concentration

T_max

Time to maximum concentration

Tl_ast

Time of last measurable concentration

CL

Clearance

V_d

Volume of distribution

T_1/2

Half-life

AUC

Area under the curve

F_u,p

Fraction unbound in plasma