(M1130-05-29) Optimization and Application of an In Vitro Direct Contact Neurovascular Unit Model to Assess Neurotoxicity

Purpose: Traditional in vitro neuronal and blood-brain barrier (BBB) models often fail to capture the complex interplay between neuroactivity and BBB integrity, resulting in poor translatability to in vivo neurotoxicity outcomes. The disruption of the neurovascular unit (NVU) by environmental neurotoxins poses serious risks to central nervous system health. Understanding how these toxins impair BBB function and neuronal viability is essential for developing effective neuroprotective strategies and identifying recovery mechanisms. To address this need, we developed and optimized a physiologically relevant in vitro direct contact NVU model that integrates brain microvascular endothelial cells, astrocytes, pericytes, and neuroblastoma cells (Figure 1). This model allows for concurrent assessment of barrier integrity, compound permeability, and neurotoxicity under both physiological and neurotoxicant-challenged conditions.

Methods: Primary human astrocytes, primary human pericytes, and brain microvascular endothelial cells (HBEC-5i) were seeded in direct contact onto Transwell^® inserts using a design of experiments (DOE) approach based on previous optimization studies. Barrier integrity was evaluated by transendothelial electrical resistance (TEER) and the apparent permeability coefficient (P_app) of 4 kDa FITC-dextran. SH-SY5Y neuroblastoma cells were then seeded into the basal chamber to complete the NVU model. Model validation included assessing compound permeability ranking (CNS-active drugs and neurotoxins) compared to known in vivo data and comparison with a prior triculture BBB model lacking neurons. Neuronal viability and outgrowth were assessed using imaging and viability assays to determine the influence of BBB integrity. Tight junction proteins (ZO-1, claudin-5, occludin, PECAM) and efflux transporters (P-gp, MRP family) were evaluated using immunofluorescence. The model’s sensitivity to neurotoxins was tested using sodium cyanide, with measurements of 4 kDa FITC-dextran and hypoxanthine P<sub>app to assess BBB disruption.

Results: The optimized NVU model demonstrated improved barrier integrity and neurophysiological relevance compared to the previous BBB triculture model. The P<sub>app values of 4 kDa FITC-dextran were significantly lower, indicating reduced paracellular permeability. Enhanced tight junction (ZO-1, claudin-5, occludin, PECAM) and efflux transporter (P-gp, MRP family) expression were confirmed via immunofluorescence (Figure 2). Compound permeability results aligned with established in vivo ranking data, validating the predictive capacity of the NVU system. Notably, integration of SH-SY5Y cells revealed that neuroactivity was modulated by BBB status, with neuronal viability and neurite outgrowth significantly affected by barrier integrity. Cyanide exposure significantly disrupted NVU function. FITC-dextran P<sub>app increased from 3.56 ± 0.28 × 10⁻⁶ to 4.21 ± 0.46 × 10⁻⁶ cm/s (p < 0.01, **). Similarly, hypoxanthine P<sub>app rose from 22.98 ± 0.96 × 10⁻⁶ to 32.01 ± 0.70 × 10⁻⁶ cm/s (p < 0.01, **), confirming toxin-induced increases in permeability. These results highlight the model’s ability to detect toxicant-driven BBB disruption and neurovascular dysfunction.

Conclusion: This NVU model offers a robust and physiologically relevant platform to study neurovascular interactions, compound permeability, and toxin-induced neurotoxicity. It demonstrates improved barrier integrity, functional transporter expression, and predictive compound ranking aligned with in vivo systems. The model also reveals that barrier compromise directly affects neuronal health, emphasizing its utility for mechanistic studies of neurovascular disruption. Cyanide exposure studies further confirm the model’s sensitivity and relevance for neurotoxicant risk assessment. This model holds strong potential for preclinical drug screening, mechanistic toxicology, and development of neuroprotective therapies.

Acknowledgements: This work was financially supported in part by Environmental Protection Agency OSAPE Star Grant #RD-84002701-0.

Figure 1. Schematic representation of the in vitro NVU model in comparison with an in vivo neurovascular cross-section.
SH-SY5Y neurons were seeded on the basal side of the Transwell® insert to simulate neuronal interactions with the blood-brain barrier (BBB). The apical side was seeded with brain

Figure 2. Immunofluorescence analysis of tight junction proteins and efflux transporters in direct contact triculture BBB and NVU models.
This figure illustrates the expression of tight junction-associated proteins (occludin, F-actin, claudin-5, ZO-1, and PECAM-1) and efflux transporters (MRP2 and P-gp) in both the triculture BBB (upper images) and NVU (lower images) models. The inclusion of neurons in the NVU model appears to enhance the expression of all analyzed proteins, suggesting neuron-mediated reinforcement of BBB integrity and transporter function.