Bhavatharini Arun 1, Aditya Anil Naik 1, Gauri G Rao 1
1 Department of Clinical Pharmacy, USC Alfred E. Mann School of Pharmacy and Pharmaceutical Sciences, University of Southern California ( Los Angeles, USA)
Objectives: Central nervous system (CNS) infections such as meningitis, ventriculitis, and encephalitis occur across physiologically distinct compartments where antibiotic distribution is highly compartmentalized [1]. Treatment-related decisions, however, often rely on sparse measurements from accessible sites that may poorly reflect target-site exposure. Entry of intravenous (IV) antibiotics into the CNS is regulated by the blood-brain barrier (BBB), which restricts paracellular passage through tight-junctions (~7-10 Å in size) [2]. Hydrophilic IV antibiotics are highly sensitive to inflammation-driven tight-junction-disruption [3]. Inflammation, e.g., elevated cerebrospinal fluid (CSF) interleukin-6 (IL-6), disrupts tight junctions and dynamically increases CNS penetration. IL-6 is a mechanistic marker of BBB permeability and disease severity due to its short half-life [4]. When systemic therapy is insufficient, intraventricular (IVT) antibiotics are used to bypass the BBB, but dose optimization remains empirical because target-site pharmacokinetics/pharmacodynamics is difficult to characterize given the invasive nature of sampling strategies [5].
Key questions:
(i) How do disease-associated changes in barrier integrity and CSF hydrodynamics shape CNS antibiotic exposure-response?
(ii) How can infection site-specific exposure-response be inferred from clinically accessible, sparsely sampled sites?
This study aimed to:
1. Develop a mechanistic CNS physiologically based pharmacokinetic (PBPK) model incorporating neuroinflammation-driven BBB permeability and CSF hydrodynamics to characterize regional CNS antibiotic exposure.
2. Integrate PBPK-predicted CNS exposure with a mechanism-based pharmacodynamic model (MBM) to quantify exposure-response relationships and optimize antibiotic regimens.
Methods:
Step-1: In vitro quantification of inflammation-driven BBB permeability: A human BBB co-culture (hCMEC/D3 endothelial cells, apical; HMC3 microglia, basolateral) was infected with CNS-derived Acinetobacter baumannii isolates to induce neuroinflammation. Barrier integrity was quantified longitudinally using transendothelial electrical resistance (TEER), and cytokine measurements up to 32 h. Paracellular permeability was quantified using sodium-fluorescein permeability, alongside apparent permeability coefficient (P_app) measurements of antibiotics. Confocal-imaging qualitatively confirmed tight-junction-protein disruption.
Step-2: Mechanistic CNS-PBPK model development incorporating infection-dependent pathophysiology: In vitro P_app values were partitioned into transcellular and paracellular components; paracellular transport was governed by a Renkin-pore-hindrance framework. Infection-driven BBB leakiness was implemented as IL-6-dependent reductions in TEER, with sodium-fluorescein constraining paracellular permeability. Additional pathophysiological determinants of CNS drug-transport (pH-dependent partitioning; CSF-hydrodynamics) were incorporated in a CNS-PBPK model comprising plasma, brain extracellular-fluid (ECF), intracellular-space, and ventricular, cisternal, and subarachnoid-space (SAS) CSF compartments. Drug-transport was represented by bidirectional-diffusion and bulk CSF-flow between compartments. Model performance was evaluated using clinical CNS exposure data: cerebral microdialysis measurements after IV meropenem dosing in brain-ECF and ventricular-CSF [6], and ventricular-CSF concentration profiles following IVT colistin [7].
Step-3: PBPK-driven CNS exposure-response modeling and dosage optimization: PBPK-predicted ventricular-exposures representative of ventriculitis informed static concentration-time kill studies to map pharmacodynamic (PD) responses under clinically expected exposure variability. IV meropenem regimens (2 g q8h infused over 2 h; 6 g/day continuous infusion; 15 g/day continuous infusion) and IVT colistin regimens (5 mg q24h; 10 mg q24h), with/without external ventricular drainage (EVD), were evaluated. A MBM with capacity limited growth, antibiotic mediated killing, and heterogeneous subpopulations was linked to PBPK simulated CNS kinetics. Candidate regimens were evaluated in the hollow fiber infection model (HFIM) for proof of concept validation of predicted antibacterial response under clinically relevant ventriculitis exposure conditions.
Results:
BBB disruption and permeability. CNS-derived Acinetobacter baumannii caused rapid barrier injury: TEER reduced by ~50% within 8 h and pro-inflammatory cytokines rose over 32 h (IL-6: 5,276 vs 279 pg/mL; IL-8: 6,317 vs 239 pg/mL; TNF-α: 42.6 vs 6.51 pg/mL). Sodium-fluorescein P_app values increased from 3.66 × 10-5 to 7.54 × 10-4 cm/s (20.6-fold), and meropenem P_app increased from 1.58 × 10-5 to 6.36 × 10-5 cm/s (4.03-fold). Renkin analysis predicted an expansion of BBB pore radius (~7 Å to ~13 Å), corresponding to a ~1.85-fold increase in the effective BBB surface area available for paracellular transport.
IV meropenem: neuroinflammation, site heterogeneity, and timing. The CNS PBPK model reproduced clinical meropenem CNS PK in brain-ECF and CSF, and revealed pronounced spatial and temporal heterogeneity. During infection, relative CNS drug exposures followed: brain ECF >cranial-SAS-CSF > spinal SAS-CSF >ventricular-CSF, consistent with better prognosis observed in meningitis than ventriculitis [8], where ventricular “sink condition” limits exposure. Time-to-peak-concentration was delayed in CNS (plasma~1 h vs ECF~2.4 h, cranial SAS~3.9 h, spinal SAS~5.1 h), and CNS elimination half-life was prolonged (~5-7 h) vs. plasma (~2 h). Consequently, single paired plasma-CSF measurements at non-optimal times may misrepresent clinically relevant target-site exposures. IL-6-mediated barrier disruption increased the meropenem exposure by ~1.7-fold in brain-ECF and ~2-fold in ventricular CSF. Conversely, CSF-acidification (pH7.3 → 7.0) decreased CNS exposure by ~40%, consistent with increased unionized drug-fraction and enhanced efflux across CNS-barriers.
IVT colistin: CSF hydrodynamics dominate. Ventricular colistin exposure profiles were accurately captured using physiologically-informed ventricular CSF-volumes (age and hydrocephalus status) and CSF turnover (intracranial pressure, CSF outflow resistance and CSF drainage). Increased CSF outflow resistance, as in infection [9], doubled ventricular-exposure, whereas EVD reduced ventricular colistin elimination half-life (~7 h to ~4 h). Following IVT dosing, exposures followed: ventricular CSF> cranial SAS-CSF> spinal SAS-CSF> brain ECF, with brain ECF remaining lowest due to limited diffusion.
PBPK-linked MBM simulations and HFIM validation. Model predicted ventricular exposures for IVT colistin and IV meropenem matched PD outcomes observed in the HFIM against carbapenem-susceptible (ATCC 19003) and resistant (Ab77) Acinetobacter baumannii strains. High-dose continuous infusion meropenem achieved effective killing in the susceptible strain, whereas intermittent dosing failed. Colistin monotherapy was ineffective due to heteroresistant subpopulations. Combination therapy consistently resulted in sustained bacterial eradication, mechanistically captured as colistin-mediated outer membrane permeabilization reducing the meropenem-EC50.
Conclusion: An integrated, biomarker-anchored framework using in vitro BBB impairment data to drive a mechanistic CNS PBPK model enabled site-specific prediction of antibiotic exposure and informed mechanism-based modeling of bacterial response in CNS infections. This framework quantitatively links neuroinflammation, CNS drug disposition, and antibacterial effect, providing a generalizable translational platform for other antibiotics and CNS pathogens. It also serves as a valuable translational tool for refining future in vivo CNS microdialysis studies in CNS infections in alignment with the 3Rs principles. This model-based approach reflects Lewis Sheiner’s pioneering vision of pharmacometrics, integrating physiology, biomarkers, and quantitative modeling to enable precision dosing and therapeutic decision making in challenging diseases.
References:
[1]. Goodlet KJ, Nailor MD. Central nervous system infections and antibiotic selection: all infection sites are not created equal. Clin Infect Dis. 2024 Aug 15;79(2):573–574. doi:10.1093/cid/ciad772
[2]. Sarin H. Physiologic upper limits of pore size of different blood capillary types and another perspective on the dual pore theory of microvascular permeability. J Angiogenes Res. 2010 Aug 11;2:14. doi:10.1186/2040-2384-2-14
[3]. Zhang S, Gan L, Cao F, Wang H, Gong P, Ma C, et al. The barrier and interface mechanisms of the brain barrier, and brain drug delivery. Brain Res Bull. 2022;190:69–83. doi:10.1016/j.brainresbull.2022.09.017
[4]. Wicha SG, Kinast C, Münchow M, et al. Meropenem pharmacokinetics in cerebrospinal fluid: comparing intermittent and continuous infusion strategies in critically ill patients. Antimicrob Agents Chemother. 2024;68(9):e00451-24. doi:10.1128/AAC.00451-24
[5]. Tunkel AR, Hasbun R, Bhimraj A, Byers K, Kaplan SL, Scheld WM, et al. 2017 Infectious Diseases Society of America’s clinical practice guidelines for healthcare-associated ventriculitis and meningitis. Clin Infect Dis. 2017 Mar 15;64(6):e34–e65. doi:10.1093/cid/ciw861
[6]. Hosmann A, Ritscher L, Burgmann H, et al. Meropenem concentrations in brain tissue of neurointensive care patients exceed CSF levels. J Antimicrob Chemother. 2021 Nov;76(11):2914–2922. doi:10.1093/jac/dkab286
[7]. Imberti R, Cusato M, Accetta G, et al. Pharmacokinetics of colistin in cerebrospinal fluid after intraventricular administration of colistin methanesulfonate. Antimicrob Agents Chemother. 2012 Aug;56(8):4416–4421. doi:10.1128/AAC.00231-12
[8]. Luque-Paz D, Revest M, Eugène F, et al. Ventriculitis: a severe complication of central nervous system infections. Open Forum Infect Dis. 2021 Jun;8(6):ofab216. doi:10.1093/ofid/ofab216
[9]. Tariq A, Aguilar-Salinas P, Hanel RA, Naval N, Chmayssani M. The role of ICP monitoring in meningitis. Neurosurg Focus. 2017 Nov;43(5):E7. doi:10.3171/2017.8.FOCUS17419
Reference: PAGE 34 (2026) Abstr 12220 [www.page-meeting.org/?abstract=12220]
Poster: Oral: Lewis Sheiner