Development of pharmaceutical treatments for neurological disease is stifled by high cost and low success rate in clinical trials. A more predictive human pre-clinical model would significantly reduce the barrier to entry while accelerating the drug discovery pipeline. Microphysiological systems (MPS) are in vitro model systems with in vivo-like physiology potentially filling this need. We have developed a novel MPS human Nerve-on-a-Chip, or NerveSim®, comprised of an iPSC-derived neuron and primary human Schwann cell 3D coculture system that forms a morphological and electrophysiological simulacrum of a human nerve. Neuronal processes are grown down a channel containing embedded electrodes capable of stimulating and recording from 10 locations along the nerve. This system is employed in a 24-well format to provide high-throughput electrophysiological characterization amenable to testing multi-drug panels for neurotoxicity on a clinically relevant nerve model. We have also developed a live-cell automated imaging analysis technique to detect neurotoxicity via morphological changes in human NerveSim® over time. Here, we tested several chemotherapeutic compounds with known neurotoxicities by correlating changes in compound action potentials (CAPs) and axonal morphology for multidimensional non-invasive longitudinal functional characterization.
CAPs are evoked with a series of current steps (1 µA, 5 µA, 10 µA, 24 µA, and 48 µA) from 6 different electrode locations sequentially in the NerveSim® channel while recording from 10 electrodes simultaneously. The average of 12 trials at each current step for all 45 electrode combinations is converted from the time domain to the velocity domain and the envelope is calculated to produce a purely positive signal. The velocity envelope signals for each current step are aligned in velocity and the maximal velocity projection (MVP) is calculated to collapse the 45 individual recordings into a single measurement. The area under the curve (AUC) of
the MVP is calculated to produce the velocity density index (VDI) metric that is then compared, on a well-by-well basis, before and after dosing with chemotherapeutics to measure functional changes from neurotoxicity. To measure morphological changes associated with dosing, we took images through the dose week and performed automated segmentation to quantify fiber length, fiber number, and object number producing a nerve degeneration index (NDI). IC50’s were then calculated for each assay type using appropriate positive and negative controls to give a multidimensional assessment of toxicity. Chemotherapeutic compounds were selected from auristatins, maytansinoids, vinca alkaloids, and antibody drug conjugates (ADCs) thereof.
Of the auristatins, MMAF was less cytotoxic than MMAE producing an IC50 approximately three thousand-fold higher as measured by electrophysiology (62nM and 0.02 nM respectively). The NDI, derived from fiber length analysis, showed IC50s with a similar trend between MMAF and MMAE but with lower sensitivity than electrophysiology (150 nM and 4.3 nM respectively). This growth metric estimates the amount of fragmentation as a reduction in fiber length. Visually, MMAF and MMAE produced clear fragmentation in human NerveSim®. The ADC HER2-MMAE DAR4 proved less toxic than its payload alone with an electrophysiology IC50 of 17,524 ng/mL and an imaging IC50 of 24,161 ng/mL. These IC50s contrast with literature reports testing N87 cancer cells where HER2-MMAE inhibits proliferation with an IC50 of 95 ng/mL. The antibody maytansine conjugate (AMC), trastuzumab-DM1 (T-DM1), and the AMC metabolite, DM4-SMe, both showed potent neurotoxicity (103 ng/mL and 0.93 nM respectively) as determined by electrophysiology. However, while the IC50 from the growth assay for DM4 SMe agreed with the electrophysiology IC50 (3.84 nM), treatment of NerveSim® cultures with T-DM1 resulted in relatively little visible fragmentation and an IC50 of 27,542 ng/mL. This growth IC50 for T-DM1 was above the maximum dose tested whereas electrophysiology detected toxicity near the lowest dose. Together, this suggests morphology was relatively unaffected in contrast to inducing significant functional detriments. The vinca alkaloids, vincristine and vinblastine, were also potent neurotoxic compounds when assayed by electrophysiology (IC50s of 1.43 nM and 7.58 nM respectively). They produced clearly visible neuronal fragmentation and growth IC50s
of similar trend to electrophysiology (IC50s of 7.68 nM and 45 nM respectively). This data supports the utility of human NerveSim® in testing novel ADCs for off-target effects on the nervous system while highlighting the necessity for functional assessment.
The NerveSim® platform is a valuable pre-clinical model for sensitively testing novel therapeutics for potential functional and morphological effects on the peripheral nervous system. Our 24-well NerveSim® EEA plate is 1,200-fold more efficient than field recordings, enabling generation of high through-put functional data in combination with novel imaging metrics to investigate multi-drug panels for safety and efficacy on human cells.
