Pharmacokinetics of bedaquiline in cerebrospinal fluid (CSF) in patients with pulmonary tuberculosis (TB)

Abstract

Background

With current treatment options most patients with CNS TB develop severe disability or die. Drug-resistant tuberculous meningitis is nearly uniformly fatal. Novel treatment strategies are needed. Bedaquiline, a potent anti-TB drug, has been reported to be absent from CSF in a single report.

Objectives

To explore the pharmacokinetics of bedaquiline and its M2 metabolite in the CSF of patients with pulmonary TB.

Patients and methods

Individuals with rifampicin-resistant pulmonary TB established on a 24 week course of treatment with bedaquiline underwent a lumbar puncture along with multiple blood sample collections over 24 h for CSF and plasma pharmacokinetic assessment, respectively. To capture the expected low bedaquiline and M2 concentrations (due to high protein binding in plasma) we optimized CSF collection and storage methods in vitro before concentrations were quantified via liquid chromatography with tandem MS.

Results

Seven male participants were enrolled, two with HIV coinfection. Using LoBind^® tubes lined with a 5% BSA solution, bedaquiline and M2 could be accurately measured in CSF. Bedaquiline and M2 were present in all patients at all timepoints at concentrations similar to the estimated unbound fractions in plasma.

Conclusions

Bedaquiline and M2 penetrate freely into the CSF of pulmonary TB patients with a presumably intact blood–brain barrier. Clinical studies are urgently needed to determine whether bedaquiline can contribute meaningfully to the treatment of CNS TB.

Introduction

CNS TB is the most deadly form of TB, constituting 1%–2% of the estimated 10 million TB cases annually.¹ Current treatment recommendations are based on pulmonary TB regimens, with little innovation or addition of new agents in decades, and limited attention to the abilities of component drugs to penetrate into the brain and CSF where Mycobacterium tuberculosis bacilli reside in the CNS. Outcomes are poor and result in severe disability or death for most.² The burden of drug-resistant (DR) CNS TB, which is almost universally fatal, is likely underestimated due to insensitive diagnostics in CSF.³ Outcomes may be improved by introducing novel antibiotics or by dose optimization of conventional anti-TB agents.

Evidence supporting the development and repurposing of anti-TB agents, such as bedaquiline, delamanid, pretomanid and linezolid, for CNS TB has lagged behind that for pulmonary TB. Bedaquiline, approved by the FDA in 2012, has had a dramatic impact on pulmonary DR TB outcomes and is now a critical component of recommended regimens for pulmonary DR TB.^4–8 The evidence for or against the use of bedaquiline in CNS TB is scanty. Bedaquiline was found to penetrate into healthy rodent brain tissue sufficiently well to achieve therapeutic concentrations.^9,10 Only a few clinical cases have been reported. Bedaquiline was undetectable in the CSF of one patient with CNS TB,¹¹ whose outcome was not reported. On the other hand, bedaquiline, along with other agents, was added to a strengthened regimen in two patients failing DR CNS TB treatment, with both patients experiencing clinical improvement.¹² The most intriguing case is that of a patient with extensively resistant pulmonary TB who participated in a clinical trial of bedaquiline, pretomanid and linezolid. Investigation of seizures that commenced after successful completion of treatment revealed a sterile tuberculoma on excisional biopsy, with confirmed resistance on molecular testing, suggesting that the regimen had treated the DR CNS TB.¹³ This handful of anecdotes remains the only published evidence on the use of bedaquiline for treatment of CNS TB.

The aim of this exploratory study was to measure total concentrations of bedaquiline and its M2 metabolite (which has 3–6-fold lower activity than bedaquiline) in CSF and to describe their plasma and CSF pharmacokinetic (PK) profiles at steady state in patients with pulmonary TB. A bioanalytical method with lower limits of quantification (LLOQ) of <1% of bedaquiline and M2 concentrations in plasma, reflecting estimated unbound drug, was needed to measure CSF concentrations. We optimized sampling and bioanalytical methods to capture the expected low bedaquiline and M2 concentrations that result from their high protein binding in plasma (estimated to be >99.9% and 99.7%, respectively) and their propensity to stick to collection devices and tubing.¹⁴ We hypothesized that only unbound bedaquiline and M2 would penetrate into the CSF.

Patients and methods

Ethics

The Human Research Ethics Committee of the University of Cape Town approved the study (HREC REF: 279/2018).

Study design

This nested exploratory study was conducted between October 2018 and January 2019 at two clinical research sites in Cape Town, South Africa (TASK Brooklyn Chest and TASK Clinical Research Centre). Participants were adults with rifampicin-resistant pulmonary TB, without CNS TB, recruited from the parent ACTG DELIBERATE trial (ACTG A5343; ClinicalTrials.gov number = NCT02583048), which investigated the effect of bedaquiline and delamanid on the QT interval.¹⁵ Participants underwent a separate informed consent process for the substudy and were enrolled if they had received 14 days of 400 mg of oral bedaquiline, followed by 200 mg three times per week until their week 24 study visit (together with multidrug background therapy), and were excluded if they had any contraindication to lumbar puncture or if they had an international normalized ratio (INR) >1.4 or platelets <100 × 10⁹/L.

Collection and storage of samples

Blood for plasma PK was collected in the context of the parent study at pre-dose and at 3 (if receiving delamanid), 5, 7, 10 and 22 h post-dose into EDTA tubes and centrifuged at 800–1000 g for 10 min. Plasma was pipetted into 1 mL aliquots and frozen below −70°C within 1 h of collection. Participants in the CSF substudy were allocated to a CSF sampling timepoint that coincided with one of the intensive plasma PK sampling timepoints. A single lumbar puncture was performed at 5, 7 or 10 h, with an atraumatic Sprotte 22 or 24 gauge needle. Lumbar puncture was performed first, followed by blood collection, to minimize the interval between samples. Four aliquots of 1 mL of CSF were collected directly into either polypropylene or LoBind^® Eppendorf sampling tubes (Table 1) and transported in a crushed ice bath with no further manipulation and frozen below −70°C within 30 min of collection and stored. The collection methods were carefully designed in light of bedaquiline’s physiochemical properties and known propensity to stick to collection and storage devices (e.g. needle for lumbar puncture, sampling tubes, pipettes and storage tubes).¹⁶

Table 1.

Effect of LoBind^® tube and BSA on analyte yield: sample storage vial investigation

	Sample tube 1, reference	Sample tube 2	Sample tube 3	Sample tube 4
Sample tube composition	polypropylene cryovial	LoBind® Eppendorf tube	LoBind® Eppendorf tube lined with a 5% BSA solution	LoBind® Eppendorf tube containing CSF sample:5% BSA solution (1:1, v/v)
Change in bedaquiline yield (range)	-	−4.3% to 8.3%	11.0% to 11.8%	17.8% to 26.4%
Change in M2 yield (range)	-	10.0% to 22.4%	120.4% to 136.1%	114.8% to 136.0%

A positive change indicates less analyte adsorption.

Sample size and statistical analysis

This was an exploratory study without formal sample size calculations. A target of two or three participants per timepoint was considered sufficient to generate preliminary data regarding CSF penetration and concentrations over the dosing interval. Individual bedaquiline and M2 concentrations were measured and summarized descriptively by sampling time. Unbound bedaquiline and M2 were estimated from the total concentration based on protein binding expected to be >99.9% and 99.7%, respectively. These results are shown graphically with mean concentration and range using Phoenix^® WinNonlin^® version 8.3 (Certara USA, Inc., Princeton, NJ, USA).

Bedaquiline and M2 CSF assay

The plasma assay for bedaquiline and M2 used in this study has been described previously.¹⁷ To optimize quantification of the bedaquiline and M2 concentrations in CSF, which we anticipated to be low, we investigated four collection and storage methods for the lowest analyte loss (Table 1). Peak area ratios of bedaquiline and M2 over their internal standards, obtained from CSF samples spiked at relatively high (1.6 ng/mL) and low (0.075 ng/mL) concentrations and stored in the reference vial (polypropylene cryovial), were compared with peak area ratios obtained from samples stored in the experimental LoBind^® tubes.

An HPLC method coupled with tandem MS detection was developed and validated for the quantification of bedaquiline and M2 in 100 μL of human CSF, in accordance with the FDA and EMA guidelines^18–20 at the Division of Clinical Pharmacology, University of Cape Town. Assay methodology is available as Supplementary data at JAC Online.

Results

CSF sampling methodology

The LoBind^® tubes with an equal volume of CSF and 5% BSA solution showed the greatest reduction in adsorption for bedaquiline and M2 collectively. However, the M2 internal standard ionization was suppressed. The LoBind^® tubes lined with a 5% BSA solution resulted in the lowest analyte loss without affecting ionization and were used for sample collection and storage, as well as the preparation and storage of calibration standards and quality controls used for analysis.

Participants

Seven male participants with a mean age of 44 years (range = 34–55) and normal serum creatinine and albumin were included (Table S1, available as Supplementary data at JAC Online). Two (29%) were living with HIV and virally suppressed on a dolutegravir-based regimen (CD4 = 253 and 558 cells/mm³, respectively). Drug regimens included bedaquiline and at least four additional drugs, such as delamanid, linezolid, isoniazid, ethambutol, pyrazinamide and terizidone.

PK results

There were 38 plasma samples and 7 CSF samples that were collected and analysed (Table S1 and Figure S1). Total plasma concentrations of both analytes were in the expected range for all participants, with an observed T_max of 5 and 7 h for bedaquiline and M2, respectively. Bedaquiline had a mean C_max of 1368.1 ng/mL and a mean AUC_last of 19 825.9 ng·h/mL. M2 in plasma reached a mean C_max of 217.3 ng/mL and a mean AUC_last of 4134.6 ng·h/mL.

In CSF, the observed C_max and T_max for bedaquiline were 3.790 ng/mL and 5 h. The concentrations ranged from 0.446 to 3.790 ng/mL across all timepoints. The observed C_max and T_max for M2 were 1.400 ng/mL and 5 h, with concentrations ranging from 0.138 to 1.400 ng/mL. Only one participant contributed a sample to the 10 h CSF PK, which was also the observed minimum concentration. The mean CSF to plasma concentration ratio was 0.12% for bedaquiline and 0.3% for M2; these CSF concentrations were similar to the estimated plasma unbound fraction of bedaquiline (<0.1%) and M2 (0.3%) (Figure 1).

Figure 1.

Mean (range) total plasma and CSF concentration profiles of bedaquiline and M2. The mean concentration and range of bedaquiline and M2 in plasma (filled and open circles, respectively) and CSF (filled and open squares, respectively) are shown for the seven participants. Estimated plasma free fractions of each analyte are displayed with pale broken lines. BDQ, bedaquiline.

Discussion

Bedaquiline and M2 were present in the CSF in all patients at all timepoints in the presence of a presumably intact blood–brain barrier (Figure 1). For any drug, the unbound drug fraction is the pharmacologically active component. Bedaquiline and M2 are highly protein bound in plasma, estimated to be >99.9% and 99.7%, respectively¹⁴ and therefore little unbound drug is available to penetrate into the site of action. CSF concentrations were expectedly much lower than in plasma, but were remarkably similar to the estimated plasma unbound fraction of bedaquiline (<0.1%) and M2 (0.3%). We estimate the unbound CSF:plasma ratio for both bedaquiline and M2 to approximate 1:1, suggesting that unbound drug is able to penetrate freely into the CSF.

In a previous case report, bedaquiline was not detected in the CSF collected from a patient with DR CNS TB.¹¹ However, for this highly protein-bound, ‘sticky’ drug, collection materials and analysis methods are critical to ensure low concentrations can be measured accurately and that the drug does not adhere to collection, storage and bioanalysis equipment. We put a number of measures in place and selected a low-enough limit of quantification for the bioanalysis assays. This resulted in CSF collection directly from the needle and the use of low-binding tubes lined with BSA to reduce adsorption and maximize analyte yield.¹⁶ The LLOQ of bedaquiline was set three orders of magnitude lower than previously and lower than the expected lowest concentration in CSF.

TB in the CNS may be found in the CSF, meninges and brain parenchyma. Drugs intended for use in CNS TB should ideally penetrate all areas in sufficient concentrations to sterilize the CNS. CSF is an imperfect substitute for CNS concentrations, with rapid turnover and little fluid mixing between cranial and caudal segments.²¹ In addition, the blood–brain and blood–CSF barriers are similar, but not identical. Bedaquiline brain tissue, but not CSF, concentrations have been reported in animal studies, limiting our ability to directly compare our findings. Brain tissue concentrations may exceed those of the CSF, as seen in preclinical studies for meropenem and delamanid, but tissue concentrations are difficult to measure in humans and would require biopsy.^12,22 Bedaquiline has a high lipophilicity and may preferentially accumulate in brain tissue, with consequent higher concentrations, rather than in the low-protein CSF matrix, though data on this are lacking.¹⁰ Drug concentrations in brain extracellular fluid can be measured in humans, but only through invasive methods, such as microdialysis.^23,24 Once an agent has entered the CSF it is expected to diffuse freely, albeit somewhat heterogeneously, throughout the CNS structures.²⁵

The study had several limitations. The sample size was small, composed of only males and included participants without CNS TB. The unbound drug fractions in CSF were not directly measured in this study. Protein binding may vary by the method used and may be inconsistent for highly bound drugs; the estimated free fraction in plasma may therefore be over- or underestimated and affect the relative proportion of bedaquiline and M2 found in CSF.^14,26 Of concern is that the CSF concentrations we observed do not reach the MIC of bedaquiline (>30 ng/mL). Bedaquiline MIC is significantly impacted by protein, with an up to 433-fold increase in MIC in a protein-rich environment.²⁷ Estimated unbound bedaquiline in plasma is substantially lower than the MIC of 30 ng/mL, yet demonstrates high bactericidal killing alone and in combination with other agents at recommended doses.^4,28 It is unclear if this MIC is appropriate in the low-protein CSF environment.

Evidence supporting the use of bedaquiline in CNS TB is currently limited to a handful of preclinical and PK studies, as well as a few case reports.^9–13 This study has determined that bedaquiline does penetrate into the CNS, is measurable and is found at concentrations similar to the estimated unbound fraction in plasma. With other TB agents, CSF concentrations are higher with inflamed meninges and often lessen as the inflammation recovers.²⁹ Even in the absence of inflammation it is useful clinical information that bedaquiline and M2 are able to penetrate through an intact blood–brain barrier. The drug must be present at the site of action in the CNS both early in treatment when the blood–brain barrier is leaky and later once the initial inflammation has resolved.

The role of bedaquiline in a CNS TB regimen still needs to be established. Clinical studies are needed to determine whether bedaquiline can contribute meaningfully to CNS TB regimens. In the interim, the question of whether bedaquiline should be withheld from patients with DR CNS TB should be revisited.

Funding

Research reported in this publication was supported by the National Institute of Allergy and Infectious Diseases (NIAID) of the National Institutes of Health (NIH) under Award Numbers UM1 AI068634, UM1 AI068636 and UM1 AI106701. C.M.U. is supported by TASK, PanACEA consortium (TRIA2015-1102) and Radboud University Medical Centre. K.E.D. is supported by K24AI150349 (NIAID, NIH).

Transparency declarations

None to declare.

Author contributions

C.M.U., C.I.S., L.W., K.E.D. and G.M. designed the study. C.M.U. and A.H.D. enrolled and managed the participants. C.I.S. and L.W. analysed the samples. C.M.U. and C.I.S. wrote the first draft of the manuscript. All authors revised the manuscript. All authors read and approved the final version for submission.

Disclaimer

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Supplementary data

Supplementary data, including assay methodology, Table S1 and Figure S1, are available at JAC Online.