A translational site-of-action PK-PD modeling platform to propose treatment shortening regimens for tuberculosis
Jacqueline P. Ernest (1), Natasha Strydom (1), Anna Fochesato (1,2,3), Maria Garcia-Cremades Mira (1,4), Véronique Dartois (5), Radojka M. Savic (1)
(1) Department of Bioengineering and Therapeutic Sciences, University of California, San Francisco; (2) Fondazione The Microsoft Research – University of Trento Centre for Computational and Systems Biology (COSBI), Italy; (3) Department of Mathematics, University of Trento, Italy (4) Department of Pharmaceutics and Food Technology, School of Pharmacy, Complutense University of Madrid, Madrid, Spain, (5) Center for Discovery and Innovation, Hackensack Meridian Health, Nutley, NJ
Objectives: To treat tuberculosis (TB), drugs must travel from the central circulation to bacterial lesions [1]. It is thus important to not only measure drug concentrations in blood, but also in lesions to understand a drug’s efficacy and efficacious dose. Clinically, patients with advanced pathology, including the presence of cavities and caseum (a cheese-like necrosis at cores of advanced granulomas) are more likely to have an unfavorable outcome compared to those without [2]. We posit that the reasons for this disjunction fall into two broad pharmacological categories. First, bacteria in caseum are largely non-replicating and more tolerant to therapeutics than fast-replicating bacteria [3], a pharmacodynamic (PD) factor. Second, drugs have difficulty reaching these avascular regions, a pharmacokinetic (PK) factor. Accordingly, in a cohort of patients with TB, cavities and caseum had substantially lower exposure than plasma in central circulation [4]. Recent successes in clinical trials indicate that site-of-disease PK-PD metrics drive treatment shortening.
However, direct measurements of site-of-disease PK are invasive and clinical studies are thus limited [5-7]. A non-primate animal model allows for rich sampling at the site of disease and for systematic assessment of drugs, but no model has been validated to be predictive of clinical site-of-action PK. The rabbit model of active TB has been proposed as a suitable model organism as it mimics the heterogeneous lesions of TB patients, particularly its mixed cellular and necrotic, caseating lesion types [8]. To bridge the gap between rabbit and human PK profiles, we developed a translational site-of-action PK model of drug penetration in rabbit and human lesions. Our overarching goal is to prioritize treatment-shortening drug regimens with optimal site-of disease PK-PD coverage for clinical evaluation.
To this aim, we have 1) validated a nonclinical animal model as a surrogate for human lesion samples, 2) implemented a modeling platform to assess and compare priority drugs, and 3) linked to outcomes in patients with and without cavitary disease.
Methods: Seventeen drugs were evaluated using the translational modeling platform. To match available clinical lesion data [4], we generated rabbit lesion PK profiles for seven drugs (rifampicin, pyrazinamide, linezolid, moxifloxacin, clofazimine, isoniazid, and kanamycin) in five distinct lung compartments (normal lung, cellular lesion, caseous lesion, caseum, and cavity wall). An aggregate of 1,213 plasma and 2,111 lesion samples from 173 rabbits were modeled. Physiological plasma-to-lesion PK models were developed for each drug using NONMEM 7.5.1 through PsN 5.3.0 where the magnitude (partition coefficient) and rate were estimated for each lesion as previously described [9]. The lesion parameters quantified in rabbits were compared to lesion parameters estimated from the clinical lung resection study to assess between-species portability of these parameters.
For 10 drugs in or entering clinical development, dose-ranging experiments in uninfected rabbits were performed to identify the dose reproducing exposure at the clinical dose or projected efficacious dose in patients. PK studies in rabbits infected with TB were performed for up to 4 weeks. Plasma was serially sampled, and tissue was collected at necropsy. Lesion parameters were estimated, and clinical simulations were performed with integration of Phase I data where available. The results were compared to various in vitro assays that replicated the various lesion microenvironments including minimum inhibitory concentration, macrophage IC90, and caseum MBC90 (the concentration that kills 1 log of non-replicating persisters in ex vivo caseum). Site-of-disease coverage was defined as PK above the relevant target for each lesion type.
To determine whether site-of-action PK-PD could be predictive of clinical outcome, the translational platform was used to simulate the site-of-disease PK-PD coverage of regimens that have entered Phase 3 trials. A coverage profile was generated for patients with and without cavities for four regimens in Study 31/A5349 and Nix-TB.
Results: Plasma PK and target-site PK models for lung, cellular lesion, and caseum were successfully developed for 17 drugs. We found that partition coefficients obtained from the rabbit and clinical data are highly correlated (Pearson correlation: r=0.92, p-value=7.6e-12), indicating that extent of partitioning is similar between species. The high degree of correlation was consistent for all lesion types. The relationship was best described by a log-log linear model. Our results indicate that clinical lesion-centric drug penetration and PKs in patients can be predicted and quantified using a rabbit model of TB.
Of the 17 validation and priority drugs investigated, only four drugs achieved full coverage in all compartments – plasma, uninvolved lung, cellular lesion, and caseum. Of all four compartments, the caseum compartment was least likely to achieve a therapeutic concentration across all drugs, owing to both the decreased PK and generally higher PD thresholds. Seven of 17 drugs were able to maintain therapeutic concentrations in plasma, lung, and cellular lesions for the entire treatment duration. Five drugs had partial coverage (coverage less than entirety of dosing interval) in plasma, lung, and cellular lesions. One drug, kanamycin, failed to achieve coverage in any compartment except plasma.
Based on analysis of Study 31 and Nix-TB regimens, the first two regimens achieving treatment shortening in four decades, we propose a rule of thumb stating that regimens with at least two drugs achieving therapeutic concentrations in caseum are associated with treatment success in patients with cavitary TB. We demonstrate the importance of site-of-action PK-PD metrics in late-stage clinical trials of TB where a distinction between regimens that cover all sites of action perform better than those that only have favorable metrics in the central compartment. These findings define a regimen profile that can cure the hardest to treat, and therefore, have a higher likelihood of achieving superior outcomes in shorter durations in late-stage clinical trials.
Conclusions: Drug exposure at the site of action has been proposed as one the top three pillars of pharmacology [10]. Here, we have validated a translational site-of-action modeling platform that can systematically assess site-of-disease drug exposure to prioritize regimens with the highest potential for treatment shortening. We have observed a high rabbit-to-human correlation, validating the rabbit model as a surrogate for costly and invasive clinical studies. Modeling the extent of penetration in rabbits can be used to predict PK-PD at the site of disease for approved drugs and clinical candidates, propose combinations and doses for clinical trials, and significantly expedite drug development.
References:
[1] Dartois, V., The path of anti-tuberculosis drugs: from blood to lesions to mycobacterial cells. Nat Rev Microbiol, 2014. 12(3): p. 159-67.
[2] Imperial, M.Z., et al., A patient-level pooled analysis of treatment-shortening regimens for drug-susceptible pulmonary tuberculosis. Nat Med, 2018. 24(11): p. 1708-1715.
[3] Sarathy, J.P., et al., Extreme Drug Tolerance of Mycobacterium tuberculosis in Caseum. Antimicrob Agents Chemother, 2018. 62(2).
[4] Strydom, N., et al., Tuberculosis drugs' distribution and emergence of resistance in patient's lung lesions: A mechanistic model and tool for regimen and dose optimization. PLoS Med, 2019. 16(4): p. e1002773.
[5] Prideaux, B., et al., The association between sterilizing activity and drug distribution into tuberculosis lesions. Nat Med, 2015. 21(10): p. 1223-7.
[6] Dheda, K., et al., Spatial Network Mapping of Pulmonary Multidrug-Resistant Tuberculosis Cavities Using RNA Sequencing. Am J Respir Crit Care Med, 2019. 200(3): p. 370-380.
[7] Kempker, R.R., et al., Lung Tissue Concentrations of Pyrazinamide among Patients with Drug-Resistant Pulmonary Tuberculosis. Antimicrob Agents Chemother, 2017. 61(6).
[8] Ernest, J.P., et al., Lesion penetration and activity limit the utility of second-line injectable agents in pulmonary tuberculosis. Antimicrob Agents Chemother, 2021: p. Aac0050621.
[9] Kjellsson, M.C., et al., Pharmacokinetic evaluation of the penetration of antituberculosis agents in rabbit pulmonary lesions. Antimicrob Agents Chemother, 2012. 56(1): p. 446-57.
[10] Wu, S.S., et al., Reviving an R&D pipeline: a step change in the Phase II success rate. Drug Discov Today, 2021. 26(2): p. 308-314.
Acknowledgment:
We would like to acknowledge the Bill and Melinda Gates Foundation for funding of this work under the grant (INV-0002483).
Disclosure:
Nothing to disclose.