Doxorubicin impairs the blood-brain barrier in a microfluidic model of the human neurovascular unit

Up to 75% of cancer patients experience neurological symptoms of cancer-related cognitive impairment, commonly known as “chemobrain”, which persist long-term in a subset of patients. The etiology is not well understood, but blood-brain barrier (BBB) disruption has been proposed as a mechanism. Rodent models used to study effects of common cancer therapies on the BBB are not easily translated to humans due to interspecies variation in expression of tight junction proteins and transporters, as well as differences in glial morphology and morphological diversity. Conversely, quantifying subtle changes to BBB permeability in human patients using standard image-based methods is challenging due to low signal-to-noise ratios. To assess the impact of a commonly used anti-neoplastic agent, doxorubicin, on a physiologically relevant model of the human neurovascular unit (NVU), we employed a commercially available brain-on-a-chip (BOC). The BOC comprises two microfluidic channels to emulate the NVU; one channel was seeded with human primary brain microvascular endothelial cells to represent the vascular compartment, and the other was seeded with human induced pluripotent stem cell microglia and primary pericytes, astrocytes, and cortical neurons to represent the brain compartment. A porous membrane separates the channels. After the cells reached confluence and established a barrier between the channels, the vascular compartment was treated as follows: a) doxorubicin 0 ng/mL (control, n=6), b) doxorubicin 600 ng/mL (representative concentration observed in patient serum immediately after injection, n=6), c) doxorubicin 1800 ng/mL (n=6), d) doxorubicin 3600 ng/mL (n=3), and e) doxorubicin 6000 ng/mL (n=3) over 2 days. Two tracers, 0.4 kDa lucifer yellow and 3 kDa dextran, were added to the vascular media to assess barrier permeability based on the ratio of tracer concentrations in the outflow from the two channels. By day 2, our preliminary data indicates that doxorubicin at 600 ng/mL produces a significant increase of permeability (~1.4-fold, p<0.001 to 0.4 kDa lucifer yellow and ~1.5-fold, p=0.002 to 3 kDa dextran) compared to control. Higher doses of doxorubicin likewise resulted in increased permeability. There was a dose-dependent increase in lactate dehydrogenase concentration (measured in subset) in the brain channel outflow, indicative of neurotoxicity. These data provide strong evidence that doxorubicin can damage the human BBB, leading to neurodegeneration. This “chemobrain-on-a-chip” model warrants further optimization to examine how doxorubicin impairs the NVU through interconnected pathways rather than a single dominant mechanism, and to investigate the potential for therapies to prevent/reduce chemotherapy-induced BBB disruption and subsequent effects. Future studies will include a complete morphological and functional characterization of each cellular component of the NVU.