Nick van der Munnik, Drew MacCannell, Michael Monine, Ivan Nestorov
Biogen, Clinical Pharmacology and Pharmacometrics
Introduction: In recent years, adeno-associated viruses (AAV’s) have demonstrated significant progress and promise as vectors for gene therapy. Although the field has continued to make great strides in developing this modality into viable therapeutics, the pharmacokinetics (PK) of these biological nanoparticles remains poorly understood. Due to the body’s response to AAV exposure in the form of the production of neutralizing antibodies, AAV’s cannot typically be administered more than once. The lack of fundamental understanding of AAV PK coupled with their unique dosing paradigm presents a significant medical need as, in many practical cases, there is only one opportunity to administer a dose optimally.
Objectives:
- Develop a physiologically-based pharmacokinetic (PBPK) model for the distribution of the transgene DNA of an AAV therapeutic targeting CNS tissue
Methods: Data was collected and collated from several studies featuring the administration of an AAV therapeutic molecule to non-human primates via intravenous (IV), lumbar puncture (LP) or intra-cisterna magna (ICM) routes. AAV DNA transgene content was quantified in blood, liver, brain, cervical, thoracic and lumbar regions of the spinal cord and lumbar and cisterna magna regions of the CSF using qPCR. These data, measured in either gene copies / ug DNA or gene copies / ul, were converted into estimates of total amount present in compartments of interest via cell count or volume estimates, respectively.
The structural model was initially optimized by iteratively performing parameter estimation with Monolix for different structures. Upon convergence to a prototype structural model characterized by an adequate balance between physiological plausibility and quality of fit, a more thorough assessment of parameter space was performed.
The algorithm, implemented in R, featured an initial round of fitting with Monolix with 500 sets of initial estimates. Initial parameter values were sampled from a log uniform distribution between 0.01 and 100. Fit parameter values were then plotted versus AIC to qualitatively inspect for clustering of these values. Based on the criteria previously described, candidate fits representative of distinct clusters in the space of parameter fits were identified. These candidate fits were further refined by another assessment featuring a tighter parameter initialization centered about the fit parameter values from the previous round.
This algorithm was used to make refinements to the model structure and to endeavor to optimize parameters for that model.
Results:
The final model structure included ten compartments with sampling from all mentioned matrices constituting eight of these compartments. The five tissue compartments featured substructure including equilibrating and persisting subcompartments. The equilibrating subcompartments have transfer mechanisms allowing for equilibration with the distributive compartments in which they are connected as well as elimination and transduction mechanisms. Transgene in the persisting subcompartments enter via a transduction mechanism from the equilibrating subcompartment. Values for transfer rate constants are listed in the table below:
| Transfer Rate Constant | Description | Value (/Day) | RSE (%) |
| k_18 | Upper CSF to Brain | 10.9 | 23.8 |
| k_81 | Brain to Upper CSF | 9.88 | 25.3 |
| k_34 | Blood to Liver | 0.266 | 25 |
| k_43 | Liver to Blood | 0.00685 | 47.5 |
| k_cs | CSF to Spinal Cord | 0.0152 | 16 |
| k_sc | Spinal Cord to CSF | 0.539 | 5.3 |
| k_cb | CSF to Blood | 6.29 | 11.8 |
| k_bc | Blood to CSF | 0.00151 | 67.9 |
| k_dn | Downward in CSF | 333 | 20.6 |
| k_up | Upward in CSF | 354 | 19.6 |
| k | Elimination from Blood | 3.98 | 33 |
| k_elim | Elimination from Tissue | 0.0281 | 32.4 |
| k_trans | Transduction in Tissue | 0.00796 | 18.2 |
| k_x3 | Peripheral to Blood | 0.00525 | 35.8 |
| k_3x | Blood to Peripheral | 0.157 | 53.1 |
This model constitutes a balance of physiologically plausible structure and simulation behavior with AIC and relative standard errors for the parameters. The model predicts rapid equilibration of the CSF compartments as well as the phenomenon of a lesser fraction of dose reaching the spinal cord tissue via IV relative to ICM and LP administrations, consistent with expectations and observations.
Conclusions: This unprecedented PBPK modeling effort for AAV gene therapy can be used to simulate AAV concentrations in the blood, CSF, liver and CNS tissues for common routes of administration (IV, LP, ICM). This capacity to predict AAV transgene concentrations at the site of action could be useful in human clinical trials that use AAV’s for CNS disorders where only blood and CSF measurements are accessible.
Reference: PAGE () Abstr 9398 [www.page-meeting.org/?abstract=9398]
Poster: Oral: Drug/Disease Modelling