(K-433) Leveraging in silico and in vitro modeling for optimization of intrathecal AAV9 delivery to the brain: validation in nonhuman primates and translation to Rett syndrome human model phenotype

Targeted, consistent, and safe delivery of genetic therapies to the brain is critical for effective treatment of CNS disorders. Intracisterna magna delivery has been applied but has safety and clinical scalability concerns. Therefore, Alcyone has leveraged use of novel in silico and in vitro modeling to develop a lumbar-access automated intrathecal catheter infusion system (Falcon) that is designed to be safe and clinically-scalable. Additionally, preclinical testing of intrathecal PKPD plays a critical role in CNS drug research and development, yet there are no available validated tools for quantitative prediction of CSF-system wide solute transport in nonhuman primates (NHP) or transformation of results to humans.  The current study a) applies novel in silico and in vitro trials to predict and optimize AAV9 solute transport to the brain, b) validates model predictions in NHPs based on PKPD distribution of AAV9-CB-eGFP administered by Falcon compared to standard lumbar puncture (LP), and c) utilizes modeling to predict AAV9 solute transport in the world’s first CNS disease-specific human phenotype of Rett syndrome and quantitatively compare model predictions to NHPs. 

Animal-specific NHP models of CSF-system wide solute transport were constructed.  In brief, CSF geometry and flow was collected in a NHP by MRI.  These measurements defined the in silico 3D computational fluid dynamics (CFD) simulation and in vitro 3D printed model and utilized to simulate and optimize spatial-temporal solute PK over a 3h period following a) standard LP and b) infusion protocol via Falcon., both with 1mL drug volume.  To validate model predictions, ten seronegative (< 1:50) female NHPs (3-6 yrs, 4.2±1.9 kg) were co-infused with gadolinium and AAV9-CB-GFP via LP (n=5) and Falcon (n=5). Necropsy and brain harvest occurred ~21 days post-dose. Brain slices were separated for IHC and biochemistry analyses.  PK results were compared in terms of a) spatial-temporal solute transport evolution and b) comparison of gadolinium % of mM dose concentration in the brain. PD results were compared in terms of IHC visual inspection, cell count, and % area stained, and VGC/DG. Following model validation, CNS disease-specific phenotypic in silico and in vitro models of Rett syndrome were constructed and tested to compare human to NHP PK.

Modeling agreed with in vivo PK neuraxial distribution with Falcon rapidly reaching the cranial space 1-5 min post-injection and LP requiring 30-60 min.  In vivo and in silico solute showed strong similarity with in vivo in terms of time-course evolution of tracer throughout the CSF system.  Quantitatively, PK of gad % dose 30-min post-injection was significantly greater with Falcon vs. LP throughout the cortical gray matter (2.83±2.30% vs. 0.57±0.74%, p = 0.008, based on N=160 ROIs) and basal ganglia (0.17±0.07% vs. 0.04±0.03, p = 0.004, N=12 ROIs) (median±STD).  In vivo PD showed Falcon^TM VGC/DG was from 2 to 16X greater than LP in brain regions (p=0.001, N=7 ROIs), and Falcon^TM VGC/DG was 4X smaller than LP in the liver (p=0.016).  The injection protocol was then simulated in the Rett in silico model and compared to PK model prediction in NHP. 

The in silico CFD and in vitro 3D printed model predictions were validated by in vivo CSF system-wide solute transport evolution in NHPs for specific intrathecal injection scenarios.  Falcon^TM, utilizing a safe and clinically-scalable approach, showed significantly improved gad PK post-injection and higher rates of viral transduction throughout the gray matter and detargeting of liver.