Within the neurovascular unit (NVU), the blood brain barrier (BBB) is composed of specialized endothelial cells that separate the brain tissues from the blood system (Fig. 2a). In order to target specific areas of the brain, drugs traveling through the blood vessels must cross this barrier. Studying the BBB is therefore key to understand brain drug targeting and efficiency.
In vivo techniques are still the best way to study the BBB. They reflect the full BBB complex environment and its diversity. However, 80% of drugs successfully tested on animals fail the following clinical testing. This points out the necessity to develop alternative solutions to costly, time-consuming, and ethically contentious animal drug testing. On the other hand, in vitro Petri dishes solutions, though offering an easy model suitable for large-scale drug testing, do not reflect the complexity of the BBB and are not suitable to address complex scientific questions.
To date, few “brain-on-chip” solutions reflecting the BBB system have been published. Van der Helm and colleagues reviewed these solutions in a paper published in Tissue Barrier in 2016 [1].

1. Booth and Kim (2012 [2], fig. 2b-I): a micro-BBB model. This double-layer PDMS micro-device integrates a polycarbonate (PC) membrane, sandwiched in two glass slides. A 4-point metallic electrode deposition (a key standard in transendothelial electrical resistance (TEER) measurement) allowed the author to characterize the relevance of their system.
2. Yeon et al. (2012 [3], fig. 2b-II): a permeability assay system for cerebral microvasculature. This PDMS-made device allows generating two different flows in two adjacent micro-channels leading to a pressure difference. This allowed us to trap HUVECs cells and generate a suitable barrier after 23h to test various drugs.
3. Griep et al. (2013 [4], fig. 2b-III): a BBB-on-a-chip. This PDMS based microdevice consists of two microchannels separated by a PC membrane. The top channel (coated with collagen I) hosts hCMEC/D3 cells and a barrier is formed at the membrane after 2 days.
4. Achyuta et al. (2013 [5], fig. 2b-IV): a quasi-NVU-on-a-chip. Interestingly, this PDMS-based device is composed of two independent parts that can be used separately for cell culture. Once assembled, it emulates a fairly complex NVU environment allowing advanced molecule trafficking studies across the BBB.
5. Prabhakarpandian et al. (2013 [6], fig. 2b-V): a SyM-BBB device. This PDMS-on glass system introduced the ability to simultaneously image the brain tissue compartment and the blood compartment. In this case, the communication was designed with micro-perforations in the PDMS itself.
6. Cho et al. (2015 [7], fig. 2b-VI): a 3-dimensional BBB-on-a-chip. Poly-D-lysine and collagen were used to host RBE4 rat endothelial cells allowing a barrier to be formed by two thin layers of cells at both sides of the membrane. Neutrophil migration through this barrier was then monitored with or without the addition of a chemoattractant (interleukin 8).
7. Kim et al. (2015 [8], fig. 2b-VII): a collagen-based 3D model of brain vasculature. Mouse bEnd.3 endothelial cells were cultured in 3D printed micro-tubes (a pattern including fluidic connector drills) to mimic the BBB. Fluorescence images tracking labeled dextran proved the functionality of the system and the barrier’s permeability. Mannitol assays further characterized the permeability and its recovery over time. Blood brain barrier microfluidic models
8. Brown et al. (2015 [9], fig. 2b-VIII): a full NVU-on-a-chip. This more complex PDMS-based device is composed of three layers respectively for a vascularization chamber separated from a brain chamber by a PC membrane and a third brain perfusion layer with micro-channels. As in the previous device, tests with FITC-labeled dextran could characterize the permeability of the BBB. Worthwhile to mention, the authors used induced pluripotent cells stem cells (hiPSC)-derived neurons. The TEER was also measured using a 4-point electrode system and a 33°C cold-shock induced a significant decreased in TEER.
9. Sellgren et al. (2015 [10], fig. 2b-IX): a microfluidic NVU model. Two PDMS layers were separated with channel imprints with either a PTFE or PC membrane. Astrocytes in a collagen hydrogel and then bEnd.3 cells in collagen IV-fibronectin (PTFE) or collagen I (PC) were then cultured in the device. Thanks to the entire transparency of the device, phase contrast microscopy could be used as an interesting readout of the system. After removing the astrocytes, the authors characterized the endothelial barrier with similar techniques (FITC-dextran). Shear stress of cells was nicely-investigated using immunofluorescence to follow claudin-5 gene expression. This study showed that a PTFE membrane was more suitable than a PC one for cell attachment under relevant shear stress.
10. Walter et al. (2016 [11], fig. 2b-X): a barrier-on-a-chip microfluidic system. The two authors also used this hardware design for intestinal and lung epithelial barriers. Two PDMS layers separated by a PET membrane (sealed with silicone) were sandwiched in between two glass slides. Electrodes in a 4-point sensing structure allowed TEER measurements. PDMS-based reservoirs were also integrated on the top glass slide. Either hCMEC/D3 cells or primary rat endothelial cells (co-cultured with astrocytes and pericytes) were used as cell models. Again, TEER assays and FITC-dextran/Evans-blue albumin tracking allowed the authors to characterize the barrier permeability. Confocal microscopy could also be used.