Determination of [¹¹C]Rifampin Pharmacokinetics within Mycobacterium tuberculosis-Infected Mice by Using Dynamic Positron Emission Tomography Bioimaging

Abstract

Information about intralesional pharmacokinetics (PK) and spatial distribution of tuberculosis (TB) drugs is limited and has not been used to optimize dosing recommendations for new or existing drugs. While new techniques can detect drugs and their metabolites within TB granulomas, they are invasive, rely on accurate resection of tissues, and do not capture dynamic drug distribution in the tissues of interest. In this study, we assessed the in situ distribution of ¹¹C-labeled rifampin in live, Mycobacterium tuberculosis-infected mice that develop necrotic lesions akin to human disease. Dynamic positron emission tomography (PET) imaging was performed over 60 min after injection of [¹¹C]rifampin as a microdose, standardized uptake values (SUV) were calculated, and noncompartmental analysis was used to estimate PK parameters in compartments of interest. [¹¹C]rifampin was rapidly distributed to all parts of the body and quickly localized to the liver. Areas under the concentration-time curve for the first 60 min (AUC_0–60) in infected and uninfected mice were similar for liver, blood, and brain compartments (P > 0.53) and were uniformly low in brain (10 to 20% of blood values). However, lower concentrations were noted in necrotic lung tissues of infected mice than in healthy lungs (P = 0.03). Ex vivo two-dimensional matrix-assisted laser desorption ionization (MALDI) imaging confirmed restricted penetration of rifampin into necrotic lung lesions. Noninvasive bioimaging can be used to assess the distribution of drugs into compartments of interest, with potential applications for TB drug regimen development.

INTRODUCTION

Tuberculosis (TB) treatment regimens were developed largely by empiricism and brute force trial and error, and only recently have pharmacokinetic/pharmacodynamic (PK/PD) assessments and mathematical modeling been used to advance drug and regimen development for TB (1–3). To optimize treatment for any infection, effective drug concentrations must be present in the compartments where the pathogens reside. Indeed, a growing number of studies support the importance of monitoring drug concentrations in infected tissues (4, 5), and tissue drug distribution studies at infected and uninfected sites are currently recommended by the U.S. Food and Drug Administration (FDA) (4). Serious consequences of inadequate drug concentration in target tissues include treatment failure and selection pressure for antibiotic-resistant organisms (6). Conversely, for some drugs, high concentrations in a tissue of interest can be associated with serious toxicities (7). Direct measurement of drug concentrations in the lung lesions where Mycobacterium tuberculosis resides can be achieved only by necropsy in animal experiments (and is limited to one time point per animal) and cannot be achieved in humans except under extreme circumstances when lung resection is planned for clinical reasons. Therefore, methods that can quantitatively assess drug concentrations at the site of infection over time are needed.

New techniques such as matrix-assisted laser desorption ionization (MALDI) mass spectrometry can be used to examine the spatial distribution of TB drugs in the lung and in TB granulomas (8). Unfortunately, these techniques are invasive, rely on accurate resection of tissues, and are only semiquantitative. To overcome these limitations, we have developed positron emission tomography (PET) as a noninvasive and real-time alternative to yield in situ PK data in live animals. In this study, we utilized C3HeB/FeJ (so-called “Kramnik”) mice, which develop a diverse range of TB lesions, including necrotic and hypoxic granulomas (9–13). Noninvasive dynamic [¹¹C]rifampin PET imaging was used to simultaneously measure rifampin (a first-line TB drug) concentrations in multiple compartments of interest over time in live animals. Ex vivo two-dimensional MALDI imaging of the lung tissues was used to provide further information about spatial drug distribution of the drug.

MATERIALS AND METHODS

All protocols were approved by the Johns Hopkins Biosafety, Radiation Safety, and Animal Care and Use Committees.

Animal infection.

Four- to six-week-old female C3HeB/FeJ (Jackson Laboratory, Bar Harbor, ME) mice were aerosol infected with frozen stocks of M. tuberculosis H37Rv, using the Middlebrook inhalation exposure system (Glas-Col, Terre Haute, IN). At least 5 mice per group were sacrificed to determine the bacillary burden as CFU, 1 day after infection and at the time of imaging, as described previously (13, 14). Separate groups of identically infected mice were used for imaging and postmortem MALDI studies.

Imaging.

[¹¹C]rifampin was synthesized at the Johns Hopkins PET Center using a modification of the methods described by Liu et al. (15). Briefly, [¹¹C]rifampin was synthesized by standard solution chemistry using [¹¹C]methyl triflate produced using the Tracerlab FX MEI system (GE Healthcare, Laurel, MD) starting from cyclotron-produced [¹¹C]CO₂. The [¹¹C]methyl triflate was bubbled through precursor solution (2 mg in 100 μl of methyl ethyl ketone) in a stream of helium (20 ml/min). The reaction was allowed to go for 5 min at room temperature and then quenched with 1.5 ml of 100 mM ammonium formate. Purification was completed by a semiprep high-pressure liquid chromatography (HPLC) Luna C₁₈(2) column (Phenomenex, Torrance, CA). Specific activities were 278.06 ± 79.95 GBq/μmol with a radiochemical purity of >99%. Live animals were imaged within in-house-developed sealed biocontainment devices with an intravenous catheter system that allows “on-table” injection of [¹¹C]rifampin and immediate subsequent PET imaging (16). Mice were weighed and injected intravenously with 8.44 ± 2.19 MBq (0.07 ± 0.02 ng) of [¹¹C]rifampin via the tail vein catheter. This is a standard radiation intravenous dose in mice for PET imaging (16–18). Dynamic PET imaging was performed over 60 min using the Mosaic HP PET (Philips, Bothell, WA) small animal imager according to the following acquisition protocol: 10 windows of 1 min, 5 windows of 2 min, and 8 windows of 5 min, based on existing knowledge of the PK of rifampin in mice. Imaging was performed in pairs with infected and uninfected (control) animals being scanned simultaneously. Computed tomography (CT) imaging was performed immediately after PET, using the CT component of the NanoSPECT/CT (Bioscan, Washington, DC) small animal imager. At least 5 mice were imaged for each group. Images were reconstructed and coregistered with CT images using Amide 1.0.4 (http://amide.sourceforge.net), and standardized uptake values (SUV) were computed as described previously (13, 17, 19). Using the coregistered CT images as a reference, spherical regions of interest (ROI) were manually drawn in the liver, heart (left ventricle), brain, and lung tissues (three per animal) and applied to the PET images. Since each spherical ROI contains both full and partial voxels depending on the orientation, correction for partial volume effects was applied to all imaging data using Amide. Since PET imaging is quantitative, [¹¹C]rifampin was calibrated using phantoms with known concentration (mass and radioactivity) to yield tissue concentrations in nanograms per milliliter at each time point.

Two-dimensional MALDI imaging.

Postmortem MALDI was performed in identically infected mice 1 h after administration of a single 10-mg/kg (of body weight) dose of rifampin (Sigma-Aldrich, Milwaukee, WI) intravenously. A total of four necrotic, three pneumonic, and three uninvolved tissue regions were analyzed from lungs obtained from three different M. tuberculosis-infected mice. The lungs were resected from each animal and immediately frozen either in isopentane (Fisher, Atlanta, GA) cooled by dry ice or by 2 min of exposure to liquid nitrogen vapor. MALDI mass spectrometry imaging and lesion sample preparation were performed at the Public Health Research Institute Center (Newark, NJ) as described previously (8) using a MALDI LTQ Orbitrap XL mass spectrometer (Thermo Fisher Scientific, Bremen, Germany) located in the biosafety level 3 facility. Spectra were acquired in negative ion mode with a mass window of m/z 500 to 900. The limit of detection (LOD) was 50 femtomoles or 410 ng/g for rifampin. Drug peak identities were confirmed by the presence of rifampin-specific product ions at m/z 722 and 397 in tandem mass spectrometry (MS/MS) spectra. Relative quantitation of drug levels within caseum and cellular lesion regions was performed using MSiReader (20). ROI were drawn in each image covering either cellular lesion or caseum, and the mean drug signal intensity within that region (normalized to the corresponding drug internal standard) was calculated following exportation of the ROI peak list into Excel.

Pharmacokinetic analysis.

Rifampin concentrations (nanograms per milliliter) estimated from PET SUV were analyzed using noncompartmental methods (WinNonlin Standard ver. 2.1; Pharsight Corporation, Mountain View, CA). Peak concentration (C_max) and area under the concentration-time curve for the first 60 min (AUC_0–60) for each tissue of interest were generated using the noncompartmental intravenous bolus model. A two-tailed Student t test was used to compare PK parameters and to calculate P values.

RESULTS

Mice were aerosol infected with M. tuberculosis, with a pulmonary implantation of 1.74 ± 0.11 log₁₀ CFU, 1 day after infection. Ten weeks after infection, and at the time of imaging, the pulmonary bacterial burden was 6.95 ± 0.67 log₁₀ CFU. Well-defined, necrotic granulomas with fibrosis at the rim were noted in the M. tuberculosis-infected animals (Fig. 1).

FIG 1.

Postmortem histology demonstrating necrotic pulmonary granulomas. Lung tissues obtained from mice 10 weeks after infection with Mycobacterium tuberculosis are shown. (A and B) Hematoxylin and eosin staining shows necrotic TB granulomas (arrows). Panel B is a high-power view of the inset in panel A. (C) Masson's trichrome staining demonstrates collagen staining (blue, arrow) surrounding the granuloma. (D to F) High-power views of the inset in panel B show cellular infiltrate with foamy macrophages (hematoxylin and eosin) (D), acid-fast bacilli (E), and reticulin staining (F) at the edge of the granuloma (red, arrow).

[¹¹C]rifampin PET/CT and postmortem MALDI images.

[¹¹C]rifampin PET/CT imaging of the lungs from representative mice is shown in Fig. 2. PET signal (purple) was lower in areas of consolidation (Fig. 2B) in the M. tuberculosis-infected mouse than in other areas of the lung. [¹¹C]rifampin PET signal was more evenly distributed in lungs of the uninfected mouse (Fig. 2C).

FIG 2.

Pulmonary [¹¹C]rifampin PET/CT imaging. (A) Three-dimensional CT reconstruction of a representative mouse is shown. The yellow line marks the location of transverse PET/CT images in panels B and C. (B) Transverse images from a representative mouse 10 weeks after infection with Mycobacterium tuberculosis demonstrate a much lower [¹¹C]rifampin PET signal (purple) in the areas of consolidation seen on CT (gray areas, outlined) than in uninfected areas. (C) More-uniform [¹¹C]rifampin PET signal is noted in the lung of an uninfected mouse. PET signal is also noted in the heart (H).

Postmortem MALDI was performed in identically infected mice 1 h after administration of a single dose of rifampin intravenously. Relative rifampin signals in necrotic, pneumonic, and uninvolved lung tissues were 0.0741 ± 0.0360, 0.3952 ± 0.0695, and 0.4402 ± 0.1372, respectively, confirming the lower rifampin signal in necrotic areas than in uninfected tissues. A two-dimensional MALDI heat map showing the rifampin signal from a representative M. tuberculosis-infected mouse is shown in Fig. 3.

FIG 3.

Postmortem localization of rifampin in infected lung sections by MALDI mass spectrometry imaging. Lungs from Mycobacterium tuberculosis-infected mice were processed for two-dimensional MALDI imaging 1 h after administration of 10 mg/kg of rifampin intravenously. Hematoxylin and eosin staining (A) and the corresponding two-dimensional heat map (high, red/yellow; low, blue/green) with relative rifampin signal ([M-H]⁻, m/z 821.399) imaged by MALDI mass spectrometry imaging (B) from a representative Mycobacterium tuberculosis-infected mouse are shown. Necrotic (solid line), pneumonic (dotted line), and uninvolved (dashed line) lung tissues are outlined in panel A. Limited penetration of rifampin into the necrotic caseum (arrowheads) is observed (B), with little signal from the drug noted in the center of the large necrotic lesions.

Multicompartment [¹¹C]rifampin PK in live animals.

Figure 4 shows the tissue concentrations of [¹¹C]rifampin as SUV over the 60-min postdose interval in liver, blood, brain, and lung compartments for M. tuberculosis-infected and uninfected animals. Since PET imaging is quantitative, the SUV were used to yield tissue concentrations in nanograms per milliliter and used for noncompartmental PK analyses (Table 1). There was no significant difference in the weight or injected dose between infected and uninfected animals. Similarly, there was no difference in AUC_0–60 between infected and uninfected mice in the liver, blood, or brain compartments (P > 0.53). However, drug exposures were significantly lower in necrotic lung lesions in infected mice than in healthy lungs of uninfected mice (63.76% ± 17.39%; P = 0.03). [¹¹C]rifampin concentrations in the brain were found to be 14.55% ± 1.67% of the levels found in blood.

FIG 4.

Dynamic [¹¹C]rifampin PET imaging in live animals. Standardized uptake values (SUV) for liver (A), blood (B), brain (C), and lung (D) compartments in Mycobacterium tuberculosis-infected (red) and uninfected (black) mice are shown. After a single intravenous dose, [¹¹C]rifampin rapidly localizes to the liver (A). No differences are noted in the tissue concentrations of [¹¹C]rifampin between infected and uninfected animals in the liver (A), blood (B), and brain (C) compartments. However, tissue concentrations were lower in infected than in uninfected lung tissues (D). Data are represented as medians and interquartile ranges. Five animals were used for each group.

TABLE 1.

Characteristics of study animals and results of noncompartmental pharmacokinetic analyses^a

Parameter	Mean ± SD for animal type:		P value
	Mycobacterium tuberculosis infected	Uninfected
Weight (g)	30.54 ± 1.90	34.18 ± 2.50	0.07
Injected dose (ng)	0.07 ± 0.02	0.07 ± 0.02	0.80
Injected dose (MBq)	8.61 ± 2.09	8.26 ± 2.51	0.80
Blood
Cmax (ng/ml)	0.0622 ± 0.0289	0.0591 ± 0.0447	0.90
AUC0–60 (ng · h/ml)	0.0056 ± 0.0015	0.0061 ± 0.0021	0.68
Brain
Cmax (ng/ml)	0.0086 ± 0.0085	0.0050 ± 0.0022	0.39
AUC0–60 (ng · h/ml)	0.0012 ± 0.0013	0.0008 ± 0.0025	0.53
Liver
Cmax (ng/ml)	0.0608 ± 0.0167	0.0656 ± 0.0156	0.65
AUC0–60 (ng · h/ml)	0.0583 ± 0.01406	0.0538 ± 0.0168	0.65
Lung
Cmax (ng/ml)	0.0221 ± 0.0095	0.0326 ± 0.0205	0.33
AUC0–60 (ng · h/ml)	0.0034 ± 0.0008	0.0054 ± 0.0015	0.03

^aRifampin concentrations (nanograms per milliliter) derived from PET imaging of Mycobacterium tuberculosis-infected and uninfected mice were analyzed using a noncompartmental intravenous bolus model (WinNonlin Standard). Peak concentrations (C_max) and areas under the concentration-time curve for the first 60 min (AUC_0–60) for different compartments are shown as means ± standard deviations. P values were calculated using a two-tailed Student t test. Five animals were used for each group.

A movie depicting the 60-min dynamic [¹¹C]rifampin PET scan from a representative M. tuberculosis-infected mouse is provided in the supplemental material (see Movie S1). After intravenous administration, [¹¹C]rifampin was distributed widely and rapidly localized to the liver with subsequent hepatobiliary clearance into the small intestine.

DISCUSSION

Rifamycins are an essential group of TB drugs, with potent sterilizing activity against M. tuberculosis. Preclinical as well as clinical studies have demonstrated that rifamycins have dose-dependent activity and that current dosing achieves concentrations that are on the steep part of the dose-response curve (22, 23). Rifamycins have significant potential for shortening TB treatment (21, 24, 25), and recent trials have demonstrated that within the range of doses tested clinically, higher doses of rifamycins improve microbiologic outcomes significantly (26, 27). Other clinical studies also suggest that increasing rifampin concentrations can hasten sterilization (28) in TB treatments and that low rifampin concentrations lead to acquired drug resistance (29). Furthermore, a recent randomized phase 2 clinical trial by Ruslami et al. demonstrated that intensified treatment with high-dose intravenous rifampin substantially lowered mortality (35% versus 65% in controls) without an increase in toxicity in patients with TB meningitis (30), a serious form of TB, with high morbidity and mortality (31, 32). Similarly, TB pericarditis is also a life-threatening form of TB (33), and low drug penetration is likely to play a role in the poor outcomes. Clearly, getting sufficient drug to the site of infection is essential for optimizing TB treatments.

While significant efforts are being targeted toward optimizing rifamycin dosing in TB regimens, limited data are available regarding drug penetration and concentrations at the site of infection. These data are essential to develop informed dosing decisions for rifamycins as well as other drugs. For example, a study by Kjellsson et al. utilizing postmortem analyses demonstrated that rifampin concentrations were lower in pulmonary TB lesions than in uninfected tissues in experimentally M. tuberculosis-infected rabbits (34). Moreover, rifampin has limited penetration into the brain compartment, and rifampin concentrations in the cerebrospinal fluid (CSF) remain below the MIC against M. tuberculosis after standard dosing (35, 36). Other drugs, though, have excellent penetration into lung lesions, CSF, and brain (34, 35). TB drugs must be given in combination to prevent resistance, and the design of regimens must take into account differential distribution of drugs into compartments of interest if acquired resistance is to be avoided.

Tomographic molecular imaging can be used to evaluate disease processes deep within the body, noninvasively and relatively rapidly. It is therefore not surprising that molecular imaging has powerfully augmented the investigation of various pathologies, both for research and for patient care (37). For example, while conventional imaging is already critical to the management of patients with cancer, molecular imaging can provide early detection of lesions, changes with treatments, identification of patient-specific cellular and metabolic abnormalities, and the distribution of cancer therapeutics in target tissues and throughout the body. Moreover, imaging is noninvasive, permitting longitudinal assessments in the same individual, which is a fundamental advantage over current tools, with significant potential for translation to humans. Another major advantage of imaging is that it provides a holistic, three-dimensional assessment of the whole organ or body, less likely to be limited by sampling errors. In fact, molecular imaging methods such as [¹⁸F]fluorodeoxyglucose ([¹⁸F]FDG) PET are increasingly being used to monitor treatments in humans (38–40). Further, dynamic PET can be performed at multiple time points following the delivery of a dose of radiolabeled drug, allowing for characterization of dynamic drug concentrations over time in any tissue of interest (41–43). We utilized whole-body PET bioimaging to study [¹¹C]rifampin PK in live M. tuberculosis-infected mice. Carbon-11 (half-life [t_1/2], 20.38 min) and fluorine-18 (t_1/2, 109 min) are two common PET isotopes. Given that rifampin does not have a fluorine atom, carbon-11 was chosen so that the radiolabeled atom could be introduced into the molecule without altering its chemical structure, which could affect drug distribution. In addition, the position of the radiolabel was chosen to ensure that it would be retained on the molecule even if the drug were metabolized in the liver (15). This was not the case for MALDI utilized in this study, which captured only rifampin. Last, we utilized a tail vein catheter system for safe, on-table drug delivery to M. tuberculosis-infected mice inside sealed biocontainment devices.

Consistent with other studies evaluating drug concentrations in tissues, we demonstrated that tissue concentrations of [¹¹C]rifampin were significantly lower in infected than in uninfected lung tissues (34), but we were able to do this using noninvasive techniques (PET). This was followed up by postmortem MALDI in a limited number of animals. Interestingly, PET imaging also demonstrated that rifampin concentrations in the brain were approximately 15% of the levels found in blood, which is consistent with human data (35). MALDI and PET imaging provide complementary information. While PET can provide simultaneous, multicompartment assessments deep within the body, noninvasively and relatively rapidly, invasive sampling and subsequent MALDI analyses could provide higher resolution, as well as determination of parent/metabolite ratios. Perhaps in the future, PET imaging could be used early in regimen development to get an “initial look” at the PK of drugs in the lesions or compartments where the microbes reside. This would be informative, as invasive sampling of infected brain, cardiac, or lung tissues is difficult/dangerous, and this limitation can be overcome by using PET imaging. Though distributions of a given drug in humans and animal models will differ, translational work utilizing PET imaging may allow us to bridge preclinical and clinical findings.

Our study has some limitations. Due to high specific activities, only nanogram quantities of rifampin (microdose) were injected into each mouse for PET imaging, and so doses given were much lower than those used for treatment. There have been concerns that extrapolation of PK data derived from microdosing could be unreliable and could poorly predict drug disposition of drugs when given at therapeutic doses. However, current evidence suggests otherwise, and in most cases, microdosing is an excellent predictive tool for full-dose human PK, compared with alternative methods (44–46). Though rapid and noninvasive, PET imaging is limited by its resolution, approximately 1 mm with preclinical scanners, so other techniques will have to be implemented if higher-resolution analyses are required. PET also cannot distinguish drug from its metabolite. For this study, this may be irrelevant as mice do not metabolize rifampin, but for studies of other drugs, determination of parent/metabolite ratios in tissues by invasive means may be necessary to complement PET data. We could image for only 60 min after administration of the tracer due to the limited radioactive half-life (20.38 min) of carbon-11, but this limitation can be overcome by utilizing longer-half-life isotopes used for PET imaging. Finally, the data presented in this study were derived after a single dose of rifampin and do not represent the steady-state levels attained after administration of multiple doses.

In summary, we have developed noninvasive imaging technologies to perform multicompartment in situ PK profiling of a first-line TB drug, rifampin, in live M. tuberculosis-infected animals. We demonstrated reduced penetration of rifampin into lung lesions and the brain, compartments where M. tuberculosis resides in pulmonary and central nervous system TB. PET technology overcomes some fundamental limitations of current methodologies for quantifying drugs in tissues and could be used to gather detailed preclinical data to inform decisions about drug dosing and drug combinations for TB regimens in the future. Finally, since PET imaging is readily available clinically, validated PET tracers could also be used to study drug PK in humans, thus providing a translational bridge early in drug development.