Prediction and validation of antisense oligonucleotide distribution in human central nervous system
Michael Monine(1), Ed Plowey(1), Eric David(1), Danielle Graham(1), Toby Ferguson(1), Stephanie Fradette(1), Daniel Norris(2), Ivan Nestorov(1)
1) Biogen, Inc., Cambridge, Massachusetts, USA; 2) Ionis Pharmaceuticals, Carlsbad, CA, USA
Objectives
- To develop a physiologically-based pharmacokinetic (PBPK) model of intrathecal (IT) administration and distribution of antisense oligonucleotides (ASOs) in human central nervous system (CNS).
- To qualify model predictions against human autopsy data characterizing ASO concentrations across CNS tissues.
Methods
Although IT administration of therapeutic ASOs has become an efficient method for targeting neurodegenerative and neuromuscular disorders, distribution of ASOs within the CNS remains poorly characterized. Dose projection for IT-administered ASOs in humans requires accurate estimation of exposures at target sites within the CNS. A preclinical PBPK model was recently developed to characterize the whole-body distribution of IT ASOs based on PK data from non-human primate (NHP) studies [1]. This model described time-dependent ASO concentrations in plasma, cerebrospinal fluid (CSF), spinal cord tissue (lumbar, thoracic, and cervical), brain tissue (e.g., pons, cerebellum, hippocampus and cortex), as well as liver and kidneys. Here, we present a “humanized” version of this model assuming similarity in the anatomical structure and whole-body distribution of ASO between the two species. ASO concentrations in human CNS tissues are scaled by the corresponding physiological tissue volumes. Until recently, no experimental validation of tissue exposure predictions in humans could be performed. So far, autopsy is the only way of sampling the CNS exposures in human that does not rely on imaging techniques. To qualify the human model, the predictions were compared against tissue autopsy data collected from several deceased persons with ALS treated with tofersen [2]. Autopsy samples were taken from multiple locations of the spinal cord and brain tissues and were analyzed using a hybridization enzyme-linked immunosorbent assay method. Individual model predictions were generated according to each individual’s treatment scenario including dosing of tofersen and time from last dose till death.
Results
Model comparison to the autopsy data demonstrated that most of the brain and spinal cord predictions are within ±3-fold interval of the observed tofersen concentrations, which is consistent with the thresholds empirically used to measure the success of in vitro-in vivo and other translational tools. Given a wide range of administered dose levels (20-100 mg), number of doses and time after last dose till death (1.4-18 months) across the deceased subjects, a good alignment of the model predictions with the data confirms the validity of the proposed NHP-human translational approach. Simulations also demonstrated that the model predictions were within ±3-fold interval of the observed data for CSF collected pre-dose during the treatment. Currently, the model can predict only average trend, whereas high variability of the observed tissue data cannot be yet characterized due to very limited sampling size (n=3).
Conclusions
The translational model, leveraging tofersen as an example, provides a valuable tool for generating reliable predictions of the human CNS exposures after IT administration of therapeutic ASOs. Our model simulations lend support to the dose justifications of tofersen in clinical trials. Notably, we did not attempt to fit the autopsy data. These simulations qualified the translational modeling approach that was developed based on NHP data alone before the autopsy data became available. Further research and model improvement are necessary to better reflect physiological differences affecting PK between NHP and humans. More human tissue data, as they become available, will be added to the analysis to inform further model development.
References
[1] Monine et al. J Pharmacokinet Pharmacodyn (2021) 5: 639-654
[2] Miller et al. N Engl J Med (2020) 383:109-119