Repurposing Drugs for Malaria through a Human Dose Prediction: A Case Study with Berzosertib

Abstract

Repurposing drugs whose clinical safety has been established offers a valuable approach to reduce the cost and time associated with the development of new drugs for malaria. Here, we investigate the potential to repurpose the anticancer kinase inhibitor berzosertib for the treatment of malaria, by assessing whether a predicted efficacious human dose for malaria is within the clinically established safety and tolerability profile. First, an oral dose fractionation study was performed in a Plasmodium falciparum‐infected humanized mouse model to evaluate the antimalarial pharmacokinetic‐pharmacodynamic properties of berzosertib. The activity of berzosertib was shown to be strongly dependent on the time in which the concentrations remain above the minimum parasiticidal concentration (MPC). A human physiologically based pharmacokinetic model was then developed and used to simulate the exposure‐response relationships and determine the effective human dose which fulfilled the target pharmacokinetic profile of sustained concentrations above the MPC. Our prediction indicates that six daily intravenous doses of 75 mg of berzosertib will be required for the treatment of an uncomplicated blood stage malaria infection, which is comparatively lower than the clinical maximum tolerated dose for cancer therapy, which was identified as 440 mg administered intravenously twice weekly over a 21‐day cycle for up to 18 weeks. Despite berzosertib's potential for repurposing, further preclinical development is necessary to ensure it meets the standard of current and emerging antimalarials.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

New treatment options are needed for malaria to combat drug‐resistant parasite strains. Berzosertib is a kinase inhibitor that is in Phase I and II clinical development for treatment against advanced solid tumors.

• WHAT QUESTION DID THIS STUDY ADDRESS?

Can berzosertib be repurposed for the treatment of malaria? Does the efficacious antimalarial human dose fall within the limits defined by the established clinical safety and tolerability data of berzosertib?

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This study demonstrated that berzosertib displays potential to be repurposed given that the predicted effective dose for malaria is lower and the duration of treatment is shorter compared to cancer chemotherapy.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

Repurposing cancer drugs for malaria treatment is an emerging area of research and the model‐informed workflow outlined in this study can be employed to identify and prioritize other drugs of interest. Furthermore, it provides a basis for utilizing berzosertib as a template for chemical modification to generate derivatives that are optimized for malaria.

Malaria continues to burden many parts of the world, with an estimated 282 million cases and 610,000 deaths reported in 2024. The causal agent for malaria, Plasmodium, is transferred to humans through the bite of an infected female Anopheles mosquito. The most virulent species, Plasmodium falciparum, is prevalent in the African region and accounts for approximately 99% of cases. ¹ Emergence of partial resistance towards the artemisinin component of the recommended first line artemisinin‐based combination therapy regimen is of major concern and highlights an urgency in developing effective treatments which can target drug resistant parasite strains. ² , ³ Discovering and developing a new drug is an inherently long and expensive process. Therefore, repurposing drugs from their original indication to another disease area provides a cost‐effective approach to accelerate the development of treatments with novel use, especially if the safety and tolerability in humans has already been demonstrated. ⁴ For tropical diseases like malaria, where the lack of commercial incentives has limited private sector participation in the discovery and development of new medicines, the drug repurposing approach holds much promise. ⁵ , ⁶

The Plasmodium kinome has been explored as a promising antimalarial drug target and the potential to repurpose anticancer kinase inhibitors for malaria has been recently investigated. The mammalian target of rapamycin (mTOR) inhibitor sapanisertib, which is under clinical development for chemotherapy, was shown to have dual potency towards Plasmodium phosphatidylinositol 4‐kinase type III beta (PI4Kβ) and cyclic guanosine monophosphate–dependent protein kinase (PKG), which translated to in vivo antimalarial activity in a humanized mouse model for malaria. ⁷ Protein kinases are responsible for regulating cellular signaling pathways and play a key role in cellular function. Protein kinase inhibitors are often promiscuous due to the conserved nature of the ATP‐binding site among kinases, which can lead to non‐selectivity and contribute to off‐target effects. ⁸ , ⁹ This poses a challenge in the oncology field as kinase inhibitors are often associated with adverse drug reactions, not only owing to the nature of the molecular drug target but also due to the intensive therapy required, such as high doses and long durations of treatment. ¹⁰ , ¹¹ However, we hypothesize that a comparatively lower dose coupled with a shorter length of treatment will be required for an acute blood stage malaria infection, which could potentially alleviate inhibition of the targeted human kinases, thereby reducing the toxic effects that are observed during chemotherapy. Encouragingly in a recent clinical trial, co‐administration of the anticancer drug imatinib with clinically used antimalarials has been demonstrated to increase parasite clearance without additional toxicity. ¹² Considering this, we reasoned that the potential to repurpose an anticancer kinase inhibitor for blood stage malaria can be supported by demonstrating that the free drug concentrations attained at an efficacious human dose for malaria are lower than that attained at the clinically established maximum tolerated dose (MTD) for cancer therapy.

Berzosertib, formerly known as VE822, VX‐970 and M6620, was identified from a preliminary screen of clinically relevant kinase inhibitors with potential activity towards asexual blood stage Plasmodium parasites. Berzosertib is a selective inhibitor of the essential human ataxia telangiectasia and Rad3‐related (ATR) protein kinase and is currently in Phase I and Phase II clinical development for combination therapies against advanced solid tumors. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² A clinical trial by Yap et al. established a safe and well tolerated recommended phase II dose of 240 mg/m² by intravenous (IV) infusion for once‐ or twice‐weekly administration in a 21‐day cycle, and further revealed that no dose limiting toxicities were observed at this MTD. ¹³ , ¹⁵ , ¹⁸ Additionally, in several clinical studies, patients received five to six cycles of berzosertib treatment over a duration of 15–18 weeks. ¹³ , ¹⁴ , ¹⁷ , ¹⁸ , ¹⁹ The reported common treatment‐related adverse events after berzosertib monotherapy were flushing, nausea, pruritus, headache and infusion‐related reactions. ¹³

To this end, we performed an antimalarial human dose prediction for berzosertib, using an approach previously described of integrating pharmacokinetic‐pharmacodynamic (PK‐PD) modeling and physiologically based pharmacokinetic (PBPK) modeling, which was shown retrospectively to be accurately predictive. ²³ The primary objective of our study was to elucidate the potential to repurpose berzosertib by comparing the free drug plasma concentrations at the predicted antimalarial dose to those reached at the clinical MTD of 240 mg/m² administered twice weekly in a 21‐day cycle.

We present the in vitro and in vivo asexual blood stage antimalarial efficacy of berzosertib and establish its PK‐PD relationship against the clinically relevant P. falciparum species of malaria infected in a humanized mouse model. Additionally, we developed a PBPK model for berzosertib, which was used to simulate the unbound drug plasma concentrations after IV infusion in humans and subsequently used to determine the population PK parameters. The efficacious human dose for malaria was then predicted through exposure‐response simulations with the unbound drug plasma concentrations and the unbound in vivo MPC, as determined in the P. falciparum‐infected humanized mouse model.

METHODS

Ethics statement

Animal studies were conducted at the Holistic Drug Discovery and Development (H3D) Centre Research Animal Facility, University of Cape Town (UCT). All procedures were performed in accordance with the South African National Standard (SANS 10386:2021) for The Care and Use of Animals for Scientific Purposes and guidelines from the South African Department of Health. ²⁴ , ²⁵ Ethical approval was granted by the UCT Animal Ethics Committee prior to study commencement (ethics approval reference number 021/015). Animal husbandry and experimental procedures were followed in accordance with the guidelines of the Faculty of Health Sciences Animal Ethics Committee, UCT. Water and food were supplied ad libitum throughout the study.

Mouse population pharmacokinetic‐pharmacodynamic model

A sequential PK‐PD model was developed for berzosertib using the whole blood time‐concentration and parasitemia data obtained from the dose escalation and dose fractionation studies in the P. falciparum‐infected NSG mouse model. Nonlinear mixed‐effects modeling implemented in Monolix 2024R1 software (Lixoft) was used to determine the model parameters, which were estimated using the stochastic approximation expectation maximization algorithm. ²⁶ For PK and PK‐PD model development, various structural and error models were evaluated and retained in the base model if it resulted in a statistically significant improved model fit as indicated by at least a 3.84 unit decrease in the objective function value (OFV) for nested models or corrected Bayesian information criterion (BICc) for non‐nested models. Additionally, diagnostic plots of the observed and the individual and population model predicted concentration and parasitemia values vs. time, the individual‐weighted residuals (IWRES) against time and predicted concentration and parasitemia, and the visual predictive checks (VPC) were used to evaluate the PK and PD models. The effect of weight, continuous and discrete dose, percentage hematocrit, percentage human and mouse erythrocytes and initial parasitemia were explored as covariates and retained only if the inclusion of the covariate was pharmacologically plausible and improved the fit of the PK and PK‐PD models. Concentration values that were below the lower limit of quantification (LLOQ) of 2 ng/mL were censored in the PK analysis. One‐ and two‐compartment PK models with first and zero order absorption and elimination were evaluated; and additive, exponential, proportional or combined additive and proportional error models were tested to model the residual unexplained variability. The PK parameters were assumed to have log‐normal distribution. A one‐compartment model with linear absorption and elimination was used to model the observed berzosertib concentrations over time (t) using a set of differential Eqs. ((1a), (1b), (1c)):

$$\frac{\mathrm{d}\mathit{Ad}}{\mathrm{d}t}=-\mathit{ka}*\mathit{Ad}$$ (1a)

$$\frac{\mathrm{d}\mathit{Ac}}{\mathrm{d}t}=\mathit{ka}*\mathit{Ad}-\left(\frac{\mathit{Cl}}{V_d}\right)*\mathit{Ac}$$ (1b)

$$\mathit{Cc}=\frac{\mathit{Ac}}{V_d}$$ (1c)

Ad in Eq. (1a) denotes the amount of berzosertib in the absorption compartment, whose absorption is described by the rate constant ka. Ac in Eq. (1b) represents the amount of berzosertib in the central compartment. Cc in Eq. (1c) is the predicted berzosertib whole blood concentration. The whole blood PK for berzosertib was parameterized in terms of ka, volume of distribution (V _d ) and clearance (Cl), additionally the unbound plasma volume of distribution (V _{d,unbound plasma}) and unbound plasma clearance (Cl_{unbound plasma}) were calculated by adjusting for the mouse blood to plasma ratio (B:P) and plasma protein binding (PPB). The residual unexplained variability was modeled by a proportional error model. The estimated individual PK parameters were used as regressors in the PK‐PD model, which modeled the change in parasitemia (P) over time using a direct‐effect PD structural model according to the differential Eq. (2),

$$\frac{\mathrm{d}P}{\mathrm{d}t}=K_{\mathit{\text{growth}}}*P-K_{\mathit{\text{kill}}}*\frac{C_C^H}{C_C^H+E{C}_{50}^H}*P$$ (2)

In Eq. (2), K _growth is the parasite growth rate constant, K _kill is the parasite kill rate constant, H is the Hill coefficient and EC₅₀ is the berzosertib effective concentration that produces the half maximal effect. The residual error was described by a proportional error model and the PD parameters were assumed to have log‐normal distribution. The Hill coefficient was evaluated at different fixed values of 0.25, 0.5, 1, 3, 5, 7 and 9, and the value which produced the lowest corresponding OFV and BICc values was fixed in the final PK‐PD model. The MPC was defined as the effective concentration that produces 90% of the maximal response (MPC_EC90), and was calculated using the PK‐PD model estimated EC₅₀ and H, according to Eq. (3),

$${\text{LogEC}}_{50}=\mathrm{Log}{\mathrm{EC}}_{90}-\frac{1}{H}*\mathrm{Log}\left(\frac{90}{100-90}\right)$$ (3)

The whole blood EC₅₀ and MPC_EC90 values were also reported as the unbound plasma EC₅₀ and MPC_EC90 by adjusting the whole blood values for the mouse B:P and PPB.

Pharmacokinetic‐pharmacodynamic driver of antimalarial efficacy

The PK parameter driving the antimalarial activity of berzosertib was determined by correlating the PK‐PD indices with the corresponding parasitemia on day 7 post‐infection. The population PK model predicted whole blood C_max, the area under the concentration‐time curve from 0 to 96 h (AUC_{0–96 h}), the percentage time in which the berzosertib whole blood concentrations remained above the MPC and the population PK‐PD model predicted parasitaemia values at 96 h after the first administration on day 3 were estimated. In Monolix 2024R1 software, the relationship between the antimalarial activity and each PK‐PD index, namely C_max/MPC, AUC_{0–96 h}/MPC and percentage time above MPC was modeled using Eq. (4),

$$\text{Parasitemia}=E_0-\left(\frac{1-I_{\max }*\text{PKPD index}}{\text{PKPD index}-{\mathit{IC}}_{50}}\right)$$ (4)

In Eq. (4) E ₀ is the baseline parasitemia, I _max is the maximum parasitemia and the PKPD index is either C_max/MPC, AUC_{0–96 h}/MPC or percentage time above MPC. The BICc and visual diagnostics were used to evaluate each model fit. The model which produced the lowest BICc value was selected as the best PK‐PD driver of antimalarial efficacy.

Human physiologically based pharmacokinetic model

Using PK‐Sim 11 software (Open Systems Pharmacology), a PBPK model was developed to describe the plasma PK of berzosertib in a human population after a 1 h IV infusion, as this is the route of administration of the clinical MTD. First, the PBPK model was built using the physicochemical and in vitro properties of berzosertib (Table S1 ), and then the fit was optimized using clinical data of a cancer patient population from Middleton et al. ¹⁴ To ensure the PBPK model was adequately predictable over a dose range, mean plasma concentration‐time data were extracted from published data after berzosertib monotherapy at single doses of 18, 90 and 210 mg/m², using Graph Digitizer 2.26 (GetData). During optimization the lipophilicity and microsomal half‐life of berzosertib, due to drug clearance by CYP3A4, were adjusted through a parameter identification, as these drug parameters were identified as the most impactful during the sensitivity analysis. ²⁷ Additionally, the organ/plasma partition coefficients were estimated using the Berezkhovsky calculation method. ²⁸ The model was further verified by comparing the PBPK model predicted PK parameters C_max and area under the concentration‐time curve from time 0 to infinity (AUC_0–inf) to those observed clinically. These simulations were performed in a virtual population of 100 individuals which matched the population characteristics of the clinical studies. The accuracy of the PBPK model was based on whether the predicted PK parameters were within twofold of the observed data. In addition to the plasma PK, the unbound plasma concentrations were simulated by correcting for human PPB. A population PK analysis was performed using the PBPK model simulated concentration‐time profiles and is described in detail in the Supplementary Material S1.

Human exposure‐response simulations

Human exposure‐response simulations were performed in Simulx 2024R1 software (Lixoft) by integrating the human population PK model for unbound plasma concentrations and the unbound plasma PK‐PD parameters as determined in the P. falciparum‐infected humanized mouse model, with the assumption that the free drug interacts similarly with P. falciparum in both human and NSG mouse blood. Intravenous infusions of various doses and dosage regimens were simulated in a virtual population of 1,000 individuals to determine the efficacious dose which effectively clears parasites, from an initial parasitemia of 1% to 0.01%, for a majority (85%) of the population. For practicality, the simulated doses were rounded up or down to the nearest 5 mg. The unbound plasma concentrations at the predicted antimalarial human dose and at the MTD for cancer therapy were simulated using the developed PBPK model and the fold difference in the C_max and exposure was calculated.

RESULTS

In vitro and in vivo pharmacokinetics

The physicochemical and in vitro absorption, distribution, metabolism and elimination (ADME) properties of berzosertib were determined experimentally, except for the pKa which was predicted in silico (Table S1 ). Berzosertib is a basic molecule with a pKa of 8.9 and a moderate lipophilicity. The compound had a high thermodynamic solubility in fasted state simulated intestinal fluid (FaSSIF) pH 6.5 and fed state simulated intestinal fluid (FeSSIF) pH 5.5 and was less soluble in the more acidic fasted state simulated gastric fluid (FaSSGF) pH 1.8. Berzosertib displayed moderate passive permeability in the parallel artificial membrane permeability assay (PAMPA). The apparent intrinsic clearance (Cl_int,app) was assessed in mouse, rat, and human liver microsomal preparations, and berzosertib displayed low clearance in all species.

The PK of berzosertib was investigated in healthy Balb/C mice through a single 10 mg/kg oral and 3 mg/kg IV administration of berzosertib (Figure S2 , Table S2 ). Berzosertib displayed a short half‐life of 3.5 h after IV administration, and V _d of 7.7 L/kg and a Cl of 26 mL/min/kg. The relatively higher observed in vivo Cl compared to the lower in vitro Cl_int,app could be a result of other mechanisms involved in the clearance of berzosertib, such as transporter‐mediated clearance and renal clearance. ²⁹ The oral exposure of berzosertib was suboptimal, as shown by the oral bioavailability of 31%.

Berzosertib is efficacious against asexual blood stage P. falciparum

Berzosertib displayed potent in vitro asexual blood stage antiplasmodium activity in the drug sensitive Pf3D7^0087/N9 and multidrug‐resistant lines PfDd2 and PfK1 of P. falciparum, with the total IC₅₀ (IC_50,total) and unbound Albumax IC₅₀ (IC_50,u) (Table S1 ) values ranging from 10.5 to 14 ng/mL and 4.4 to 5.9 ng/mL, respectively after a 96 h incubation with berzosertib (Figure S2 ). There was no evidence of cross‐resistance in the tested strains as displayed by the comparable activities in the drug sensitive and resistant parasite lines.

The in vivo asexual blood stage activity and PK of berzosertib was investigated in a P. falciparum‐infected humanized mouse model using ascending doses of berzosertib from 0.5 to 20 mg/kg, which were orally administered once a day for four consecutive days, starting from day 3 post‐infection (Figure 1 ). The effect on peripheral blood parasitemia of each treatment group was measured from day 3 to 7 post‐infection using flow cytometry. The 15 and 20 mg/kg dosage groups displayed comparable activities of a 99.9% reduction in parasitemia relative to the untreated group on day 7 post‐infection, and following four consecutive oral administrations of 10 mg/kg/day, a 98.5% reduction in parasitemia was observed (Figure 1 a, Table S3 ).

Figure 1.

In vivo asexual blood stage antimalarial efficacy and PK of berzosertib. (a) Dose–response and (b) whole blood concentration‐time profiles of berzosertib in Plasmodium falciparum Pf3D7^0087/N9‐infected humanized mice after four single daily oral administrations of escalating doses of 0.5–20 mg/kg of berzosertib. Data are presented as the mean (SD) from two biological replicates (n = 2).

Berzosertib exhibited dose‐dependent increases in the C_max and AUC_{0–24 h}, which were approximately dose proportional from 0.5 to 20 mg/kg (Figure S3 ). The exposure and C_max at the 20 mg/kg dosage group (AUC_{0–24 h} = 3,019 ng.h/mL, C_max = 311 ng/mL) were higher than that attained at the 15 mg/kg dosage group (AUC_{0–24 h} = 1901 ng.h/mL, C_max = 157 ng/mL); however, this did not translate to greater efficacy as a maximum effect was observed at the 15 mg/kg dosage group. The effective dose that reduced parasitemia by 90% relative to the untreated group on Day 7 post‐infection (ED₉₀) was 9.2 mg/kg/day for four consecutive days, and the corresponding C_max (C_max,ED90) and exposure (AUC_{0–24 h,ED90}) were 98 ng/mL/day and 1,051 ng.h/mL/day, respectively (Table S3 , Figure S4 ).

Pharmacokinetic‐pharmacodynamic relationship of berzosertib in P. falciparum‐infected humanized mice

To further elucidate the PK‐PD relationship of berzosertib in P. falciparum Pf3D7^0087/N9‐infected humanized mice, a dose fractionation study was performed where total doses ranging from 10 to 60 mg/kg were administered as single doses on day 3 following infection, or total doses ranging from 20 to 60 mg/kg were administered as two equally divided doses on day 3 and 5 following infection. Nonlinear mixed‐effects modeling of observed data from the dose escalation and dose fractionation studies was performed to develop a sequential PK‐PD model and to determine the PK‐PD driver of antimalarial efficacy. The diagnostic plots showed adequate agreement with the PK and PK‐PD model predicted and observed data (Figures S5 and S7 ).

A one‐compartment model with linear absorption and elimination best described the whole blood concentration‐time data of berzosertib after oral administration. Linear PK is supported by box and whisker plots of the dose normalized C_max and AUC_{0–96 h}, which displayed dose proportionality over the experimental dose range (Figure S6 ). The effect of the covariates weight, percentage hematocrit and percentage human erythrocytes were explored however, inclusion of these covariates did not significantly improve the model fit and were therefore excluded from the final model. The estimated whole blood population PK parameters ka, V _d and Cl were 0.44 /h, 1.39 L and 0.25 L/h, respectively (Table S4 ). Berzosertib f _u,p in mice was 0.09 and had a B:P of 2.06 at a 77% hematocrit (Table S1 ), which was representative of the hematocrit achieved during the treatment period due to daily human erythrocyte engraftment in humanized mice. This resulted in a V _{d,unbound plasma} of 7.50 L and a Cl_{unbound plasma} of 1.35 L/h.

A direct‐effect PK‐PD model was used to determine the typical whole blood EC₅₀ and MPC_EC90, which were estimated to be 13.24 and 18.12 ng/mL, respectively, and once adjusted for mouse B:P and f _u,p (Table S1 ), the unbound plasma EC₅₀ and unbound plasma MPC_EC90 were 0.58 and 0.79 ng/mL (Table 1 ).

Table 1.

Estimated population pharmacokinetic‐pharmacodynamic (PK‐PD) parameters for berzosertib in Plasmodium falciparum Pf3D7^0087/N9‐infected humanized mice

PK/PD parameter	Estimate	R.S.E. (%)	2.5th percentile	97.5th percentile	Shrinkage (%)
H	7	–	–	–	–
K growth (1/h)	0.038	6.65	0.033	0.043	–
K kill (1/h)	0.058	8.85	0.048	0.068	–
Whole blood EC50,blood (ng/mL)	13.24	43.7	6.16	28.44	–
Unbound plasma EC50 (ng/mL)	0.58	–	–	–	–
Whole blood MPCEC90 (ng/mL)	18.12	–	–	–	–
Unbound plasma MPCEC90 (ng/mL)	0.79	–	–	–	–
Standard deviation of the random effects
ωKgrow	0.1	62.3	0.038	0.28	−0.84
ωKkill	0.012	28.8	0.0072	0.021	1.86
ωEC50	0.62	45.2	0.28	1.37	1.46
Error model
Proportional	0.3	7.91	0.25	0.35	–

H, hill coefficient; K _growth, parasite growth rate constant; K _kill, parasite kill rate constant; EC₅₀, effective concentration that produces the half maximal effect; MPC_EC90, maximum parasiticidal concentration or effective concentration that produces 90% of the maximal response; R.S.E., relative standard error.

A nonlinear regression analysis was performed to evaluate the concentration‐ or time‐dependent PK‐PD determinant for berzosertib antimalarial efficacy by comparing the relationship between the parasitemia at 96 h after the first treatment and the whole blood PK parameters relative to the MPC, which were C_max/MPC, AUC_{0–96 h}/MPC and percentage time in which the concentration remains above the MPC. The percentage time above MPC displayed the lowest statistically significant BICc of 70 compared to the C_max/MPC (BICc = 137) and AUC_{0–96 h}/MPC (BICc = 105) (Figure 2 ). This indicates that berzosertib exhibited time‐dependent activity in the P. falciparum‐infected humanized mouse model.

Figure 2.

Pharmacokinetic‐pharmacodynamic (PK‐PD) driver of antimalarial efficacy. Nonlinear regression analysis displaying the relationship between the PK‐PD indices (a) C_max/MPC, (b) AUC_{0–96 h}/MPC and (c) percentage time above MPC and the corresponding antimalarial activity of berzosertib after a dose fractionation in Plasmodium falciparum Pf3D7^0087/N9‐infected humanized mice. Goodness of fit indicated by the corrected Bayesian information criterion (BICc).

Human physiologically based pharmacokinetic model and pharmacokinetic parameters

A human PBPK model was developed for the total plasma and unbound plasma concentration‐time profiles of berzosertib after a 1 h IV infusion, as this is the route of administration from which the clinical MTD was determined. The physicochemical and in vitro ADME properties of berzosertib (Table S1 ) were used as input parameters in the model and refined through a parameter identification and optimization using observed clinical data from Middleton et al. ¹⁴ The predicted plasma concentration‐time profiles after single 1 h IV infusions of 18, 90, and 210 mg/m² of berzosertib were in good agreement with the observed clinical data and fell within the 5 and 95% prediction interval (Figure 3 ). The PBPK model was verified using additional clinical data and the predicted C_max and AUC_0–inf were all within twofold of the observed data indicating acceptable accuracy of the PBPK model (Table 2 ). The free berzosertib plasma PK profiles were simulated using the f _u,p in humans (Table S1 ), and later used for nonlinear mixed‐effects modeling to determine the human population primary PK parameters. A two‐compartment model with linear elimination best described the concentration‐time data (Table S6 ). The diagnostic plots showed adequate agreement with the model prediction and the PBPK model simulated profiles (Figure S8 ). Additionally, adequate agreement was demonstrated between the PK parameters that were estimated through the PBPK model generated data and those estimated from two Phase I clinical studies (Terranova et al.), which included data from 240 patients (Table S6 ). ²⁷

Figure 3.

Human population physiologically based pharmacokinetic (PBPK) model simulations. Berzosertib plasma concentration‐time profiles following 1 h IV infusions of 18 mg/m² (black), 90 mg/m² (blue), and 210 mg/m² (red) in humans. Dots represent the observed clinical data from Middleton et al. ¹⁴ (n = 3 for 18 mg/m², n = 6 for 90 mg/m² and n = 2 for 210 mg/m²) and mean lines and shaded 5^th and 95^th percentiles represent the PBPK model simulated profiles (n = 100).

Table 2.

Mean physiologically based pharmacokinetic (PBPK) model predicted and observed plasma PK parameters of berzosertib following 1 h IV infusions of berzosertib in humans

References	Dose (mg/m2)	Cmax (ng/mL)			AUC0–inf (ng.h/mL)
		Predicteda	Observed	Predicted/observed	Predicteda	Observed	Predicted/observed
Middleton et al. 14	18b	101	72	1.40	565	351	1.61
Middleton et al. 14	36	201	176	1.14	1,130	913	1.24
Middleton et al. 14	60	336	345	0.97	1884	1700	1.11
Yap et al. 13	60	531	286	1.86	2,436	1,510	1.61
Middleton et al. 14	72	404	536	0.75	2,261	2,280	0.99
Middleton et al. 14	90b	505	451	1.12	2,826	2,370	1.19
Yap et al. 13	120	1,063	662	1.61	4,871	5,440	0.90
Middleton et al. 14	140	785	817	0.96	4,396	4,550	0.97
Shapiro et al. 15	140	764	652	1.17	4,087	5,670	0.72
Middleton et al. 14	210b	1,177	808	1.46	6,594	6,130	1.08
Yap et al. 13	240	2,126	1,290	1.65	9,743	6,990	1.39
Yap et al. 13	480	4,252	4,410	0.96	19,485	14,900	1.31

C_max, maximum concentration; AUC_0–inf, area under the concentration‐time curve from time 0 to infinity.

^a Virtual population: n = 100.

^b Observed clinical data used for model development.

Human antimalarial dose prediction

The efficacious antimalarial human dose of berzosertib was predicted through exposure‐response simulations with the population free plasma concentrations after IV infusion (Table S6 ) and the PK‐PD parameters as determined in the P. falciparum‐infected humanized mouse model (Table 1 ). As indicated by the PK‐PD requirement for efficacy, berzosertib displayed time‐dependent activity; therefore, the human dose was optimized to ensure the unbound plasma concentrations remained above the unbound plasma MPC_EC90 of 0.79 ng/mL until the parasites were effectively cleared, as indicated by a parasitemia level at the limit of detection of 0.01%.

The first exposure‐response simulation performed was a single IV infusion of the dose at the clinical MTD of 240 mg/m², which, based on an average body surface area of 1.84 m², is approximately equivalent to 440 mg. This dosage regimen was predicted to be ineffective as displayed by parasite proliferation at approximately 80 h, which corresponded to drug levels falling below the MPC (Figure 4 ). The administration frequency was therefore increased to maintain effective concentrations above the MPC for three parasite lifecycles, totaling 144 h. The MTD of 440 mg was administered as either 145 mg q48h, 75 mg q24h, or 35 mg q12h. The 145 mg q48h regimen was ineffective for most of the population, whereas the 75 mg q24h and 35 mg q12h regimens displayed similar effective parasite clearance beyond 0.01% parasitemia for 80‐85% of the population.

Figure 4.

Human exposure‐antimalarial response simulations. (a) The unbound plasma concentrations (Cc) (left) and (b) corresponding antimalarial effect (P) (right) after 1 h IV infusions of 440 mg (orange), 145 mg q48 h (blue) and 75 mg q24 h (red), and 35 mg q12 h (green) of berzosertib over three parasite lifecycles. Black line represents the population median and shaded areas represent 9 bands.

A comparison of the unbound plasma C_max and exposure was performed between the clinical chemotherapeutic MTD for a 21‐day cycle and the antimalarial regimens, 75 mg q24h × 6 and 35 mg q12h × 12 (Table 3 ). Relative to the MTD, the 75 mg q24h regimen displayed a 5.3‐ and 5.7‐fold lower C_max and exposure, respectively, and the 35 mg q12h regimen displayed a 10.6‐ and 6.3‐fold lower C_max and exposure, respectively. Although the 35 mg q12h regimen displayed a greater fold difference in the C_max than the 75 mg q24h regimen, both exposures were relatively comparable. Additionally, the 75 mg q24h regimen required half the number of administrations, which could lead to better patient compliance. Therefore, 75 mg q24h for six consecutive days (75 mg q24h × 6) was proposed as the efficacious antimalarial dosage regimen. When comparing the proposed effective dose for malaria to one course of cancer chemotherapy, which is up to six 21‐day cycles of treatment, the extent and duration of berzosertib exposure is significantly lower, as shown by an estimated 34‐fold lower AUC_0–inf for malaria treatment compared to the AUC_0–inf attained at the clinical MTD for one course of cancer chemotherapy. Given this, there is expected to be reduced interaction with the human target and off‐target effects, which could potentially improve safety and tolerability. ¹³ , ¹⁸

Table 3.

Unbound plasma maximum concentration and exposure differences between the clinical MTD for one cycle and one course of cancer therapy and the predicted efficacious doses for malaria treatment

PK parameter	One 21‐day cycle of cancer therapy/one course of treatmenta	6‐day treatment for malaria
	Clinical MTDb	75 mg q24h × 6	Fold differencec	35 mg q12h × 12	Fold differencec
Cmax (ng/mL)	116	22	5.3	11	10.6
AUC0–inf (ng.h/mL)	3,571/21426a	624	5.7/34a	567	6.3/37a

C_max, maximum concentration; AUC_0–inf, area under the concentration‐time curve from 0 h to infinity.

^a One course of treatment is equivalent to six 21‐day cycles.

^b 440 mg twice a week for a 21‐day cycle.

^c Fold difference = C_max or AUC at clinical MTD/C_max or AUC of antimalarial dose.

DISCUSSION

Repurposing clinically established drugs offers an attractive approach to develop therapeutics with alternative use to their original indication because, in addition to potentially reducing development costs and timelines, this strategy presents a reduced risk of failure as the safety and tolerability in humans have already been investigated. ⁶ Berzosertib is a selective inhibitor of the human ATR kinase, and additionally, has shown favorable in vitro and in vivo asexual blood stage activity against P. falciparum. The potential to repurpose this anticancer kinase inhibitor for uncomplicated malaria was explored by predicting an efficacious antimalarial human dose and comparing the free drug concentrations at the predicted dose for malaria to those attained at the clinically established MTD for cancer therapy.

The observed in vitro activity of berzosertib translated to in vivo efficacy in a P. falciparum‐infected humanized mouse model, where a 99.9% reduction in parasitemia was displayed after four single oral administrations of 15 mg/kg/day of berzosertib compared to an untreated control group and displayed an ED₉₀ of 9.2 mg/kg/day for four consecutive days. To further elucidate the PK‐PD relationship, a dose fractionation study was performed, which revealed a strong positive correlation between the PK‐PD index percentage time above the MPC and the in vivo efficacy, which indicates that the antimalarial activity of berzosertib is time‐dependent. The humanized mouse model for malaria is a valuable tool for determining the PK‐PD relationships in the clinically relevant species P. falciparum, which would otherwise be unable to grow in standard rodent models, thereby providing confidence in the translation from the preclinical to clinical stage. The absence of adverse effects at the highest efficacious dose in the humanized mouse model and the clear dose–response and PK‐PD relationships suggest adequate selectivity towards P. falciparum in mice.

The human PBPK model and population PK analysis for berzosertib after IV infusion displayed adequate agreement with the published clinical plasma PK data. The PBPK model was optimized and verified using observed clinical data from cancer patients, which is a limitation of this study. Differences in gene expression levels for drug metabolizing enzymes and drug transporters have been reported between healthy individuals and cancer patients, where, for example, increased expression of cytochrome P450 (CYP) genes has been observed within certain cancerous tissues, which may have an effect on the elimination of a drug. ³⁰ , ³¹ This concept extends to PK variations between cancer and malaria pathophysiology, which has not been accounted for as there is insufficient data available to compensate for any disease‐related differences in the disposition of berzosertib, such as the distribution and clearance. This does warrant further clinical investigation, as possible PK variations could influence the predicted antimalarial dose.

Human population exposure‐response simulations were performed with the unbound berzosertib plasma concentrations and the unbound plasma MPC_EC90 of 0.79 ng/mL. Since berzosertib displayed time‐dependent activity, the administration protocol was optimized to ensure the concentrations remained above therapeutic levels until efficacy was achieved. Multiple dosing was required for sufficient PK/PD target engagement, and the effective concentration was maintained for three 48 h parasite lifecycles, as this was simulated to result in higher parasite clearance compared to treatment over two parasite lifecycles. The higher dose administered less frequently (75 mg q24h × 6) was preferred over the lower dose administered more frequently (35 mg q12h × 12), since this is expected to result in better treatment adherence. At the predicted efficacious antimalarial human dose of 75 mg q24h × 6 administered via 1 h IV infusion, the respective simulated unbound plasma C_max and exposure were 5.3‐ and 5.7‐fold lower than the observed C_max and exposure at the MTD of 440 mg (240 mg/m²) administered twice weekly over a 21‐day cycle, as established by Yap et al. ¹³ A 6 day duration of treatment was predicted to effectively clear parasites, whereas chemotherapy requires a longer treatment period where repeated doses are administered once‐ or twice‐ weekly over a 21‐day cycle and the duration of treatment continues for five to six cycles or 15–18 weeks. ¹³ , ¹⁴ , ¹⁷ , ¹⁸ , ¹⁹ Therefore, since no dose limiting toxicities were observed in cancer patients at the clinical MTD of 440 mg, the predicted lower dose and shorter duration of antimalarial treatment is expected to reduce off‐target inhibition and alleviate prolonged engagement with the human ATR kinase, thereby potentially reducing side effects. ¹³ , ¹⁸ However, possible differences in drug tolerability between cancer patients and malaria patients will need to be evaluated in a clinical setting.

Severe malaria does see cases where drugs, such as quinine and artesunate, are administered intravenously for several days. However, this is not ideal for uncomplicated malaria where oral administration is the conventional delivery path for most antimalarials. Therefore, evaluating the feasibility of developing berzosertib for oral administration is needed. Additionally, long‐acting injectable formulations are another potential option to explore as this method allows slow drug release, thereby increasing the elimination half‐life, which would be beneficial in reducing the administration frequency.

Berzosertib has not yet been evaluated in reproductive and developmental toxicity assays and considering that malaria greatly affects children under 5 years of age and pregnant women, investigating all possible safety concerns will be a requisite step for advancement. Preclinical toxicology studies showed that berzosertib is selective towards human ATR and is unlikely to exhibit off‐target activity against other human kinases. ³² It has also been reported that compared to cancer cells, normal cells can tolerate ATR inhibition as they are able to activate compensatory DNA damage response signaling via the ataxia‐telangiectasia mutated pathway. ³³ , ³⁴ , ³⁵ Lastly, berzosertib was not genotoxic in the bacterial Ames test and showed no evidence of cardiovascular liability or hemolytic activity in human blood. ³⁶ Although these findings are encouraging, comprehensive toxicology studies relating to these special populations are needed to support progression towards clinical application for malaria. Additionally, further preclinical development of berzosertib is required to ensure that it aligns with or improves on the profiles of existing antimalarials. For example, parasite recrudescence was not accounted for during the efficacious human dose selection and was not defined in the PK‐PD model; however, this is an important factor to elucidate since it is highly associated with treatment failure. Identifying a suitable combination partner, such as a clinically relevant antimalarial, will be important in managing parasite recrudescence and the emergence of drug resistance and merits further investigation.

Although the findings presented here will need to be validated in a clinical context, berzosertib has displayed potential to be repurposed from cancer therapy to malaria treatment. The efficacious antimalarial dose is predicted to be lower, and the duration of treatment is predicted to be shorter compared to the established dosage regimens for chemotherapy. Consequently, the free drug plasma exposure is simulated to be approximately fivefold and 34‐fold below the established safety margin of the clinical MTD after one cycle and one course of anticancer therapy, respectively, which could improve safety and tolerability during malaria treatment. Additionally, by leveraging existing clinical pharmacokinetics, safety, and toxicity profiles, this approach to predict antimalarial human doses can facilitate the identification of other drug repurposing candidates and molecular scaffolds for medicinal chemistry optimization, thereby offering a promising avenue to discover and evaluate new antimalarial agents.

FUNDING

The South African Medical Research Council, Neville Isdell for the Neville Isdell Chair in African‐centric Drug Discovery and Development for which K.C. is the chair, the National Institutes of Health under award number [R01AI152092] and the Carnegie Corporation of New York are gratefully acknowledged for funding support.

CONFLICT OF INTEREST

The authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

D.R. wrote the manuscript; D.R., M.M., M.M.M., M.N., L.K., and K.C. designed the research; D.R. and M.M.M. performed the research; D.R., M.M., M.N., and L.K. analyzed the data.

Supporting information

Data S1.

Data S2.

Data S3.

DATA AVAILABILITY STATEMENT

All data associated with this manuscript are available in the main text or the supplementary materials.