This is the first study that demonstrates relationships between drug dose, route of administration and target occupancy of a bronchodilator drug. Consistent with the view that inhalation provides local enrichment of drug in lung tissue, a higher target occupancy was observed in the lungs after inhalation compared to after intravenous infusion of ipratropium, at equivalent plasma drug concentrations. Following a more detailed analysis, the plasma concentration-derived apparent affinity (Ki) for ipratropium was an order of magnitude higher after inhalation. A similar analysis of the exposure-occupancy relationship for the pituitary gland, an organ known to also be rich in muscarinic receptors, revealed no significant difference in Ki estimates between the two conditions. These results not only concur with the established understanding that inhalation leads to a therapeutic advantage in terms of maximizing the local (occupancy-driven) therapeutic effect whilst minimizing the systemically driven side-effects, but also provides a uniquely explicit quantification of such, indicating a 10-fold improvement in therapeutic window.
There is a large unmet need for biomarkers in respiratory drug development beyond forced expiratory volume. By using the muscarinic receptor system as a model system, this study demonstrates the feasibility of using PET for examination of drug-target engagement in pulmonary tissue and highlights a potential to close the knowledge gap between drug pharmacokinetics and pharmacodynamics in early clinical development. The presented methodology thus shows promise for aiding future drug development programs targeting underlying mechanisms and cure of respiratory disease.
Quantification of 11C-VC-002 binding—method development and limitations
The effect of ipratropium on radioligand binding was evaluated by pairwise PET measurements performed on the same experimental day, in which the main outcome parameter was a ratio between radioligand binding at post-dosing over baseline conditions (the VT ratio). By this experimental design, some of the noise that was introduced in a systematic fashion was anticipated to cancel out. Importantly, this provided the basis for a few key simplifications that were central for the study delivery from a practical perspective. These simplifications will be discussed in the following paragraphs.
No motion correction was performed on the PET data, which make the lungs appear spatially smoothed according to their motion through the breathing cycle. This was justified by the fact that the animals were ventilated during the study and that the lung movement was under the same rigorous control at both baseline and post-dosing conditions. The smoothing effect was thus assumed to be similar during each PET experimental session and thus cancel out.
In contrast to the brain, which is a relatively homogenous tissue, the tissue content in the lungs represents only ~ 12% of the total volume with the rest comprised predominantly of air (~ 72%) and blood (~ 12%) [
18]. Therefore, binding parameters calculated without correcting for the inhomogeneity will likely be underestimated compared to the true binding values. Importantly, inspection of transmission data and of the initial slopes of the early radioactivity peak in the lungs indicated that air and blood content, respectively, did not substantially change between baseline and post-dosing conditions, which speak to the validity of these assumptions.
The lungs are in close vicinity to other organs with high radioactivity, e.g., the liver. The challenge with emission noise from nearby high uptake organs was primarily addressed by reducing the lung ROI so that voxels in the vicinity of these organs, such as the liver and the heart, were excluded from it. This was confirmed by the lack of high radioactivity voxels in the summation PET image within the final ROI. The influence of motion and interference from neighboring high radioactivity organs was also addressed by using an inverse-variance-weighted pooling of voxel-wise lung VT estimates. The validity of this weighting scheme was supported by preliminary assessments, indicating that voxels with low VT at the edge of the lungs (i.e., with more motion) and voxels high VT closer to high radioactivity organs or containing more blood, such as near the back, tended to have higher absolute SE (and thus variance) than typical in-lung voxels, which, in contrast, tended to have rather similar SE, irrespective of VT. Various region of interest-based quantification methods were also evaluated during preliminary analyses, but proved to be overall less robust than the presented voxel-based method, yielding whole lung VT values with higher SE and higher cross-subject variation (data not shown), thus supporting the choice of the voxel-based methodology.
Radioligand metabolites in the target tissue usually complicate accurate quantification of PET data. Since neither the lungs nor the pituitary gland has physiological barriers to plasma, radiolabeled metabolites enjoy free access to the tissues. In the present study, the total plasma radioactivity concentration was used as an input for the quantification in the lungs and pituitary gland, i.e., no metabolite correction was applied. This approach was expected to provide reasonable estimates of overall total organ binding under the assumption that the parent compound and its radiolabeled metabolites have similar kinetics of non-specific binding and, furthermore, that the labeled metabolites either do not bind to the target or bind with a similar affinity as the parent compound. Furthermore, Visser et al. demonstrated that 88% of the radioactivity in rat lung tissue at 15-min post-injection was constituted by parent radioligand [
8].
The results presented herein indicate that despite all abovementioned challenges, it was possible to arrive at quantitative estimates on
11C-VC-002 binding in the non-human primate lungs and pituitary gland. Though the inclusion of additional animals on each dose point would have been beneficial from a statistical perspective, these quantitative estimates in turn provided consistent data for establishing a relationship between muscarinic receptor occupancy and plasma ipratropium concentration. Besides
Ki estimates, the fitted occupancy model also incorporated binding potential representing
11C-VC-002 binding to muscarinic receptors at baseline conditions (see Fig.
5 with BP
ND estimates in the legend). The estimated value of about 7 in the lungs is substantial when compared to typical CNS radioligands and agrees with the known high affinity of
11C-VC-002 to muscarinic receptors and the high density of these receptors in the lung tissue [
19]. Though
VT estimates are probably underestimated due to the relatively low tissue density (high air content), the BP
ND estimates were expected to be relatively unaffected by this because
VT ratios are used to derive BP
ND (see Eq. 1).
Drug delivery via inhalation
The efficiency of drug aerosol inhalation is governed by the interplay between drug formulation, drug delivery device, and the recipient lung of the animal or patient. Following inhalation, drug deposition takes place in both conducting airways and alveolar lung regions where the gas exchange occurs. Although it was difficult to determine the exact regional drug deposition in the current experimental setup, previous studies have shown that drug deposition in mechanically ventilated lungs is more central (airway) relative to that observed in consciously inhaling patients [
20]. Whilst the advantage of inhalation for separating occupancy of the lung and parotid gland was clearly demonstrated at the total receptor population levels, it is thus feasible that more proximal sub-populations of lung receptors have higher occupancy than the distal populations. Although the identity of the airways, whose caliber modulation by antimuscarinics is critically responsible for the reduction of airway resistance in COPD patients, are not clearly defined, there is empirical evidence suggesting that antimuscarinic drug deposition in the airways, rather than alveoli, is responsible for drug effect [
21]. Much to our disappointment, however, preliminary attempts at examining regional differences in drug-induced receptor occupancy were unfruitful in the current study, largely due to considerable inter-individual variability in regional
11C-VC-002 binding. In this context, it is worth mentioning that a PET study with the nebulized radiolabeled drug could increase our understanding of regional drug distribution in the NHP lung under the current conditions. Nevertheless, future PET studies in human subjects are expected to provide a greater opportunity to image the proximal lung and thus aid in the identification of potential differences in drug-induced receptor occupancy between the airways and alveoli.
Translating and applying the findings to early clinical drug development
The current study was conducted in non-human primates, and not in humans, for radiation safety purposes. Importantly, PET is a translational imaging technique, and from several perspectives, a study in human subjects may be conducted with greater ease and with more precision than the present study in non-human primates. Moreover, a greater number of blood samples can be sampled during a PET study in a human subject, and hence, the resulting blood curve usually contains considerably less noise. Finally, due to the larger size of the human lung, there is a greater opportunity to evaluate potential gradients in drug-induced muscarinic receptor occupancy between central and peripheral lung regions.