Bart van Lieshout 1, Aanisah Hanuun 1, Gabriel Forn-Cuní 2, Herman Spaink 2, Elke Krekels 1,3, Coen van Hasselt 1, Rob van Wijk 1
1 Division of Systems Pharmacology and Pharmacy, Leiden Academic Centre for Drug Research, Leiden University (Leiden, The Netherlands), 2 Animal Sciences and Health, Institute for Biology Leiden, Leiden University (Leiden, The Netherlands), 3 Certara Inc (Princeton, The United States of America)
Introduction:
Tuberculosis (TB) is a leading cause of death worldwide[1]. Translational pharmacometrics is vital to accelerate the development of novel drugs and regimens. While rodent studies can inform early predictions of clinical anti-TB drug efficacy[2], their duration, cost, and ethical burden necessitate New Approach Methodologies[3,4]. Zebrafish larvae (ZFL) provide an ethical, cost-effective, and high-throughput preclinical experimental disease model for anti-TB pharmacology, with their translational potential previously demonstrated for isoniazid[5].
Objectives:
This study aimed to strengthen the ZFL model as a preclinical experimental disease model for translational anti-TB pharmacology, using the first-line drug bedaquiline (BDQ). To this end, we characterized the pharmacokinetics (PK) and pharmacodynamics (PD) of BDQ in ZFL infected with Mycobacterium marinum (Mm) and performed translational predictions of a clinical early bactericidal activity (EBA) study.
Methods:
EXPERIMENTAL PROCEDURE: BDQ (0-50 µM) was administered to ZFL for PK and PD experiments through continuous waterborne treatment in medium at 3-5 days post-fertilization (dpf), as well as in a wash-out experiment. BDQ PK was quantified based on daily concentration measurements from medium and pooled homogenates (n=5 ZFL). For BDQ PD assessment, ZFL were inoculated with fluorescent (mWasabi) Mm at 1 dpf, and the bacterial burden was measured longitudinally via fluorescence.
PKPD MODELLING: Sequential PKPD analysis in NONMEM® (version 7.5)[6] was performed utilizing FOCE+I. BDQ PK in medium and ZFL was modelled with first-order absorption and linear elimination, assuming a volume of distribution equal to physical ZFL volumes[7]. A covariate analysis was performed with age to account for developmental changes in absorption and clearance. For the bacterial dynamics and BDQ PD, different bacterial growth and concentration-response equations[5] were assessed alongside time-varying parameters and delays for bacterial growth and BDQ effect. Inter-individual variability (IIV) was included on the baseline bacterial burden a priori and further evaluated on structural parameters. Model selection was driven by numerical (OFV, RSE) and graphical diagnostics, alongside pharmacological plausibility.
TRANSLATIONAL PREDICTIONS: A population PK model for BDQ in TB patients from literature[8] was reproduced in rxode2[9]. PK profiles of patients in a previously published clinical EBA BDQ study[10] were simulated (100-400 mg daily). Baseline bacterial burden values were fixed to individual estimates derived from an intercept-slope model using nlmixr2[11]. Several translational scaling factors were evaluated for predicting clinical EBA[10], including the minimum-inhibitory concentration (MIC) ratio between Mm and Mycobacterium tuberculosis[5,12] or bodyweights. Additionally, various exposure metrics were explored as the primary driver of the clinical effect. Prediction accuracy was evaluated against observed clinical data obtained from the TB-PACTS database[10,13] using the root-mean-squared error (RMSE) and the mean bias error (MBE).
Results:
EXPERIMENTAL DATA: 165 PK and 161 PD observations were analyzed. The PK profile showed rapid absorption and a clear dose-exposure relationship. BDQ exposure slowed bacterial growth in an exposure-dependent manner, which was delayed, and demonstrated net killing at the highest dose.
PKPD MODELLING: The final PK model included an age-dependent increase in clearance (+181% at 5 dpf; ΔOFV:-5.02) and dosage correction factors to account for BDQ adsorption to the well plates (ΔOFV:-63.6). Bacterial growth was delayed by ~48 hours (ΔOFV:-36.9) and followed a logistic growth curve (0.085 h⁻¹; ΔOFV:-16.3) and was proportionally inhibited by the ZFL BDQ concentration. This effect was characterized by an Emax equation (EC50: 26.5 µg/mL; ΔOFV:-870.9) with the maximum effect increasing from 1.09 to 1.37 after 24 hours of BDQ exposure (ΔOFV:-8.2). IIV was quantified on the growth rate constant with a coefficient of variation of 74.6%.
TRANSLATIONAL PREDICTIONS: Simulations of clinical EBA based on the ZFL exposure-response relationship identified scaling the EC50 with the MIC ratio as the most critical translational factor, showing a close alignment with the observed bacterial burden[10,13] across all doses and timepoints. MIC-scaling reduced the prediction error (MBE: -0.0535; RMSE: 0.177) compared to the base model without translational factors (MBE: 0.119; RMSE: 0.220). In contrast, substituting plasma concentrations for whole-body exposure (MBE: 0.361; RMSE: 0.489) or applying allometric scaling to potency (MBE: -0.523; RMSE: 0.701) resulted in poor predictive performance. Combinations of factors did not improve predictions over using MIC-scaling alone.
Conclusions:
Clinical EBA results of BDQ for TB can be accurately predicted based on the PKPD from the zebrafish TB disease model. This strengthens ZFL as a resource-efficient and sustainable preclinical platform for early anti-TB drug discovery and development.
References:
Acknowledgements:
We gratefully acknowledge the support of the Farmaceutische Wetenschappen Sectorplan as part of the Sectorplan Bèta-II of the Ministry of Education, Culture and Science of the Netherlands. Data used in the preparation of this abstract and presentation were obtained from the Critical Path Data and Analytics Platform, which is administered by Critical Path Institute.
References:
[1] WHO. Global Tuberculosis Report 2025. 2025. https://www.who.int/publications/i/item/9789240116924
[2] Susanto BO, et al. Translational predictions of phase 2a first-in-patient efficacy studies for antituberculosis drugs. Eur Respir J. 2023;62(2):2300165. doi:10.1183/13993003.00165-2023
[3] FDA Modernization Act 2.0, S 5002, 117th Cong (2021). https://www.congress.gov/bill/117th-congress/senate-bill/5002/text
[4] EMA. Guideline on regulatory acceptance of 3R testing approaches. 2016. https://www.ema.europa.eu/en/documents/scientific-guideline/guideline-regulatory-acceptance-3r-replacement-reduction-refinement-testing-approaches_en.pdf
[5] van Wijk RC, et al. Anti-tuberculosis effect of isoniazid scales accurately from zebrafish to humans. Br J Pharmacol. 2020;177(24):5518-5533. doi:10.1111/bph.15247
[6] Beal SL, et al. NONMEM 7.5 Users Guides. ICON plc; 2020. https://nonmem.iconplc.com/nonmem750
[7] Guo Y, et al. Three-dimensional reconstruction and measurements of zebrafish larvae from high-throughput axial-view in vivo imaging. Biomed Opt Express. 2017;8(5):2611-2623. doi:10.1364/boe.8.002611
[8] Svensson EM, et al. Population Pharmacokinetics of Bedaquiline and Metabolite M2 in Patients With Drug-Resistant Tuberculosis. CPT Pharmacometrics Syst Pharmacol. 2016;5(12):682-691. doi:10.1002/psp4.12147
[9] Wang W, et al. RxODE: Facilities for Simulating from ODE-Based Models. CPT Pharmacometrics Syst Pharmacol. 2016;5(1):3-10. doi:10.1002/psp4.12052
[10] Diacon AH, et al. Randomized dose-ranging study of the 14-day early bactericidal activity of bedaquiline (TMC207) in patients with sputum microscopy smear-positive pulmonary tuberculosis. Antimicrob Agents Chemother. 2013;57(5):2199-2203. doi:10.1128/aac.02243-12
[11] Fidler M, et al. Nonlinear Mixed-Effects Model Development and Simulation Using nlmixr and Related R Open-Source Packages. CPT Pharmacometrics Syst Pharmacol. 2019;8(9):621-633. doi:10.1002/psp4.12445
[12] Wicha SG, et al. Forecasting Clinical Dose-Response From Preclinical Studies in Tuberculosis Research. Clin Pharmacol Ther. 2018;104(6):1208-1218. doi:10.1002/cpt.1102
[13] Critical Path Institute. Tuberculosis – Platform for Aggregation of Clinical TB Studies (TB-PACTS). https://c-path.org/tools-platforms/tb-pacts/
Reference: PAGE 34 (2026) Abstr 12295 [www.page-meeting.org/?abstract=12295]
Poster: Drug/Disease Modelling - Infection