In Situ Detection of Neuroinflammation Using Multicellular 3D Neurovascular‐Unit‐on‐a‐Chip

The human neurovascular system is a complex network of blood vessels and brain cells that is essential to the proper functioning of the brain. Researchers have become increasingly interested in the system for developing drugs to treat neuroinflammation. Currently, creating neurovascular models begins with animal models, followed by testing on humans in clinical trials. However, the high number of medication failures that pass through animal testing indicates that animal models do not always reflect the outcome of human clinical trials. To overcome the challenges of the issues with animal models, a neurovascular‐unit‐on‐a‐chip system is developed to accurately replicate the in vivo human neurovascular microenvironment. By replicating the human neurovascular unit, a more accurate representation of human physiology can be achieved compared to animal models. The ability to detect proinflammatory cytokines in situ and monitor physiological changes can provide an invaluable tool for evaluating the efficacy and safety of drugs. Using nanosized graphene oxide for in situ detection of inflammatory responses is an innovative approach that can advance the field of neuroinflammation research. Overall, the developed neuroinflammation‐on‐a‐chip system has the potential to provide a more efficient and effective method for developing drugs for treating neurodegenerative diseases and other central nervous system diseases.


Introduction
Neuroinflammation has been implicated as a key factor in the development of many neurological diseases, such as multiple sclerosis, [1] Huntington's disease, [2] Parkinson's disease, [3] DOI: 10.1002/adfm.202304382Alzheimer's disease, [4] and brain tumors. [5]pecifically, it has been demonstrated that neuroinflammation contributes to the progression of these conditions by disturbing the equilibrium between proinflammatory [6] and antiinflammatory cytokines [7] and modifying neurotransmitters [8] and receptor levels. [9]urthermore, it has been reported that neuroinflammation is a major contributor to the cognitive [10] and behavioral impairments [11] that are linked to these conditions.Consequently, comprehending the underlying mechanisms of neuroinflammation is essential for devising effective therapeutic interventions.An in-depth exploration of the fundamental processes is crucial for enhancing our knowledge of the factors driving neuroinflammation and formulating more efficient treatment strategies. [12]This type of research could offer valuable insights into molecular and physiological mechanisms at play and pave the way for identifying new therapeutic targets.Accurate models of human neurovasculature will prove invaluable in this pursuit and aid in creating innovative therapies for these devastating conditions.
To this end, the advanced neurovascular microfluidic model can help us understand physiological changes in neurological diseases (Figure 1a).Surrogate animal models have been used to simulate neuroinflammation, [13] investigate neuronal pathophysiology, [14] and develop potential therapeutic drugs. [15]owever, animal models have several limitations, including high costs, long-term development and clinical trial periods, ethical considerations, and a limited ability to predict drug effects in human patients. [16]This emphasizes the need for alternative methods to evaluate drug efficacy and safety.Some studies have highlighted the significance of species-specific properties of microvascular cells, which might contribute to the differences between the human blood-brain barrier and those in other mammalian species. [17]Moreover, in vitro neurovascular models hold significant potential as valuable tools for studying cellular behavior in controlled and repeatable environments.For example, multicell-layer models using trans-wells allow for high-throughput screening of mixed-cell populations using human cells. [18]However, the cultured conditions in these models can compromise the developed state of many cell types, resulting in disruptions and anomalies in the blood-brain barrier. [19]This highlights the need to develop improved in vitro models that more accurately mimic the human neurovasculature and its cellular components.Human organs-on-chips are a promising technology for replicating organ-level activities at the single-cell level using in vitro cell/tissue culture methods.They encompass designed microscale polymeric channels that can manipulate nanoliter sample volumes. [20]For example, in terms of spatial design (e.g., 3D cultures), physiological stimuli (e.g., shear stress), and cellular variation (e.g., cocultures and intercellular communication), they can be replicated at increasing levels of complexity. [21]y constructing models with high accuracy, we can not only see biological components that are hard to recognize in the tradi-tional 2D cell culture but also quantify and manipulate the dynamics of in vivo, and thus use them for simulating human diseases and doing high-throughput screenings of therapeutic molecules.
To address the aforementioned issues, herein, we developed an innovative "neuroinflammation-on-a-chip" combined with a nano-biosensing system to i) overcome the limitations of existing 3D neural cell culture systems, ii) detect neuroinflammation signals in a sensitive and real-time manner, and iii) facilitate the progress of human organ-on-a-chip technology.
The advanced system we have designed integrates a human organ-on-chip model, effectively replicating the microenvironment of the human neurovascular unit.This is achieved through two parallel microchannels separated by a permeable, flexible membrane lined with human vascular endothelial cells, thus providing a 3D interface between vascular and neural tissues (Figure 1b).Our developed neuroinflammation-on-a-chip system comprises a microfluidic-based neural cell culture model that emulates a typical neurovascular unit in vitro.This system consists of two parallel microchannels lined with human vascular endothelial cells and seeded with astrocytes and neurons.We applied the neuroinflammation-on-a-chip system to investigate neuroinflammation by injecting lipopolysaccharide (LPS), [22] a potent inflammatory mediator, into one microchannel.[25] However, inflammatory cytokines generally serve as mediators rather than direct inflammation triggers; therefore, utilizing them alone as inducers may not represent the complexity of inflammation.Furthermore, the use of A is confined to specific conditions, like Alzheimer's disease, limiting its applicability to broader neuroinflammation contexts.[28] While LPS has been employed in both in vitro 2D cell culture and in vivo mouse models to induce inflammation, [29][30][31] to the best of our knowledge, integrating an in vitro neurovascular unit with LPS treatment has not been explored.Our developed system could also detect proinflammatory cytokines secreted by cells in response to LPS, such as tumor necrosis factor- (TNF-) and interleukin (IL)-6, [32] by continuously monitoring and measuring the secreted cytokine levels within the neural channel.5b,33] In addition, it has been established that our neuroinflammation-on-a-chip technology has a high degree of repeatability, which makes it an excellent tool for evaluating the efficacy of novel drugs and therapies targeted at treating neuroinflammatory diseases.Using human neural cells in the chip and mimicking the in vivo neurovascular microenvironment, the neuroinflammation-on-a-chip system offers a more accurate representation of human neuroinflammation than traditional animal models, reducing the need for extensive animal testing and ethical concerns.In addition, utilizing a multicellular, microfluidic-based system to quantitatively evaluate the effects of various drugs and therapies on neuroinflammation in real time provides a powerful tool and opens up new possibilities for developing effective treatments for neuroinflammatory diseases.

Fabrication of In Vitro Neurovascular-Unit (NVU)-on-a-Chip Model
The NVU can be divided into two parts: one side is a blood vessel, and another is neural tissue.A blood vessel, in particular, is composed of vascular endothelial cells, which form the ves-sel's surface, and various brain cells, such as astrocytes and neurons, which are maintained and linked to one another.Two vertically aligned channels were designed to mimic the NVU, with the upper channel representing a blood vessel and the other side representing neural tissue (Figure 1b and Figure S1 (Supporting Information)).Several types of brain cells, including astrocytes and neurons, were selected and deposited in the bottom channel of the microfluidic chip.Human induced pluripotent stem cell-derived neural progenitor cell (iPSC-NPC)-derived neurons were revealed to be on layer 3, whereas astrocytes were found to be linked to the top side of the bottom channels.After investigation, it was confirmed that both cell types were found to be in the same place, layer 2. In addition, vascular endothelial cells (human cerebral microvascular endothelial cells; HCMECs) were injected into the top channel (layer 1), which HCMECs then encircled during a ten-day period.When designing an in vitro neurovascular unit, it is imperative to consider the "communication window" between the blood vessels and brain tissue.To fulfil this requirement, a porous membrane was utilized to separate the two channels and provide a medium for exchanging signals between the channels.This arrangement is essential for the proper functioning of the system. [34]The pore size of this basement membrane is around 5 μm, which is too small to allow neural cells and vascular endothelial cells to pass through, but large enough to allow cell secretomes, such as inflammatory cytokines, to pass through (Figure S2, Supporting Information).Vascular and neuronal cells can interact and communicate signals across these microscopic channels within a single unit, allowing them to communicate with one another.The flow rate of the culture media in the upper channel was around 100 μL h −1 , resulting in shear stress comparable to that of the blood arteries in the central nervous system. [35]he flow rate in the lower channel (10 μL h −1 ) was lower than in the top channel, allowing brain cell viability to be preserved in the lower channel.These differentiated fluxes mimic the human body's NVU system.Two cross-channels were involved in the development of neuroinflammation: one at the top and one at the bottom of the channel.These devices could be used to inject external material into the brain, such as inflammatory chemicals (LPS), causing neuroinflammation.
Figure 2a depicts a NVU-on-a-chip that has been constructed and displays the locations of the horizontal and vertical channels in separate colors.Green was the color emitted from the primary (horizontal) channels, while red was the color emitted from the secondary (vertical) channels.In addition, the intersection site indicated a yellow color, which verifies the connectivity of the channels.The injected inflammation substance into the vertical channel can be blended into the main channel; the NVU comprises human vascular endothelium, astrocytes, and iPSC-NPC-derived neurons.To verify multilayered neurovascular cells in the microfluidics, iPSC-NPCs were cultured on both sides of the porous membrane and were well-attached and differentiated (Figure S3, Supporting Information).In addition, vascular endothelial cells, red-fluorescence-protein-expressed U87, and green-fluorescence-protein (GFP)-expressed U87 were cocultured and confirmed the three independent layers into the microfluidic device (Figure S4, Supporting Information).The crosssectional picture of the fabricated microfluidic chip in Figure 2b shows the porous membrane dividing the main channel into two independent channels.After seeding vascular endothelial cells, astrocytes, and neurons onto the surface of each channel, they are maintained for many days to construct a NVU onto the microfluidic chip.Additionally, the transendothelial electrical resistance (TEER) levels could provide useful information on neurovascular maturation. [36]The TEER value grew steadily for a week af-ter sowing and then plateaued at 2.8 K cm 2 (Figure 2c).That suggested that the astrocytes and vascular endothelial cells had completely covered and developed the porous membrane.This TEER value can be maintained for a month without the neurovasculature being affected by any injectable materials.Three separate cells were labeled with different staining markers to validate the established NVU model: Tight junction protein (ZO-1) for HCMECs, Glial fibrillary acidic protein (GFAP) for astrocytes, and Neuron-specific class III beta-tublin (Tuj-1) neurons.Various types of neurovascular cells could grow successfully and be maintained within the confines of a single microfluidic chip, as shown in Figure 2d.ZO-1, one of the critical proteins for neurovasculature and the blood-brain barrier (BBB) structure, was particularly well-expressed.

Emulate Neuroinflammation in the NVU-on-a-Chip
For the induction of neuroinflammation, there should be essential inflammatory substances in the NVU.LPS is one of the molecules commonly used to cause inflammation and is a wellknown representative of this type of material.If LPS is applied to the neural channel (bottom), it can stimulate astrocytes and neurons, which can then secrete proinflammatory cytokines, including TNF-, IL-1, and IL-6 (Figure 3a).They cause an increase in inflammation and have adverse effects on the brain and vascular endothelial cells, such as the disruption of tight junctions and the stimulation of cell apoptosis, which can ultimately result in cell death.To verify the neuroinflammation model, LPS was injected into the bottom channel, and the integrity of the vascular endothelium was observed to change.First, pictures of tight junctions and cell viability were taken using optical and immunostaining techniques to determine their status (Figure 3b).The typical neurovascular model had a distinct tight junction at the edge of the vascular endothelial cells, as shown in models containing a tight junction.The LPS-treated model had difficulty distinguishing between cells.Furthermore, the fluorescent image did not clearly depict the ZO-1 protein.Inflammation-induced tight junction breakdown was shown in these photos.Over 80% of neurons either did not express Tuj-1 or did so at a low level, reducing their survival (Figure 3c).These findings may show that the NVU-ona-chip has been affected by neuroinflammation.The permeability of the vascular endothelium-astrocyte layer was investigated to corroborate the BBB's breakdown due to neuroinflammation (Figure 3d).The evaluation of microcirculation and permeability was conducted by introducing fluorescein-isothiocyanate (FITC)tagged dextran to the top channel in the experiment.The measurement of the fluorescence of the solution that was collected at the outlet of both channels provided evidence that FITC-tagged dextran had moved from the upper to the lower channel.
21f,37] In other words, the TEER value reduction can serve as a neuroinflammation signal.With the nontreated chip (negative control), no fluorescence signal was measured on the bottom channel.Neuroinflammation-on-a-chip generated by LPS, on the other hand, confirmed that ≈30% of the signal compared to the injected FITC-dextran was observed in the bottom channel.It indicated that FITC-dextran translocated from the upper to the lower channel as a result of inflammation-induced tight junction breakdown.Interestingly, it was confirmed that permeability was increased even when TNF-, one of the proinflammatory cytokines, was applied to the bottom channel.This discovery demonstrated that proinflammatory cytokines can increase the permeability of the blood-brain barrier by activating both neural cells and vascular endothelial cells.Thus, it can be concluded that proinflammatory cytokines can bring about a neuroinflammatory response.Monitoring the TEER value and measuring the changes in permeability proved to be one of the most effective methods of detecting neuroinflammation in the brain (Figure 3e).As previously stated, a decrease in TEER value indicated tight junction degradation and a reduction in BBB integrity correlated with neuroinflammation progression.The inflammation substrates (TNF-, LPS) were treated to the bottom channel of the microfluidic chip on day 14, when the TEER value was stably maintained for 10 days.After applying LPS, the TEER value was drastically decreased and showed about 30% integrity compared to the untreated (normal) neurovascular state (Figure 3f).This value can be interpreted as neuroinflammation because it reflects the breakdown of the blood-brain barrier caused by the NVU-on-a-chip LPS treatment.Furthermore, the TNF-treated model showed a similar trend of decreased TEER values, corresponding to the permeability test.Likewise, while the TEER reduction and increased permeability in this group are substantially lower than in the LPS-treated group, this evidence suggests that proinflammatory cytokines might cause neuroinflammation when they are released.

In Situ Detection of Neuroinflammation Using a Reduced-Graphene-Oxide (rGO)-Functionalized Microfluidic Channel
Next, we applied our developed neuroinflammation-on-a-chip method to study the early stages of an inflammatory NVU.In this experiment, our neuroinflammation-on-a-chip approach allowed us to investigate the early stages of an inflammatory NVU in a controlled and repeatable environment.By mimicking the human neurovascular microenvironment, we could more accurately replicate the physiological changes that occur during inflammation compared to traditional animal models.Our method also enabled us to detect proinflammatory cytokines and monitor real-time physiological changes, such as barrier function.Our ability to evaluate the efficacy and safety of drugs used to treat neuroinflammation and neurodegenerative diseases has been significantly enhanced through the application of this tool.
For example, the release of proinflammatory cytokines from brain tissue resulted in a loss in BBB integrity and a decrease in TEER value.That being said, some intriguing evidence suggests that monitoring proinflammatory cytokines might be an innovative tool for detecting early brain inflammation. [38]In the developed NVU-on-a-chip, the neuroinflammation can be simulated by administering LPS, as described above.Neuroinflammation negatively influences various neural cells, including astrocytes and neurons, and simultaneously weakens the tight junctions of vascular endothelial cells. [39]The release of proinflammatory cytokines from activated brain cells triggers this inflammatory response. [40]Quantifying the real-time production of proinflammatory cytokines generated by activated brain cells enables prompt recognition of neuroinflammation. [41]In this study, specialized aptamers, which have a selective affinity for proinflammatory cytokines, were utilized to quantify the concentration of proinflammatory cytokines in real time.In detail, first, rGO was immobilized on the bottom channel, and fluorescent-dye-tagged aptamers capable of selectively binding with proinflammatory cytokines were attached to the rGO through - interactions (Figure 4a).Immobilization of rGO into the bottom channel of the chip was confirmed by the electron microscope image and the D and G bands of the Raman spectrum (Figure 4b,c).Aptamers on the rGO are usually well-attached to the rGO surface, but when the target proinflammatory cytokines come close, they detach from the rGO and bind to the proinflammatory cytokines by a competitive reaction.
The aptamer concentration can be measured using this phenomenon by collecting the cell culture medium from the bottom channel treated with LPS and measuring the fluorescence signal.By measuring the fluorescent signal during neuroinflammation, we can gain insight into the amount of proinflammatory cytokines released from neural cells, which then allows us to make estimations about the concentration of these cytokines released by neurons in the affected area.This method has the advantage of not using an immunoassay method that requires cell sacrifice and several reaction steps using antibodies and signaling enzymes.The target proinflammatory cytokines in this study were IL-6 and TNF- by measuring each fluorescence signal (IL-6:Carboxytetramethylrhodamine {TAMRA}, TNF-:FITC) that changes according to the LPS concentration.Before measuring the LPS-induced release of proinflammatory cytokines, IL-6 and TNF- were quantified using an aptamer-rGO sensing platform with cell-free configuration, ranging from 0.125 to 4.00 μg mL −1 (Figure S5, Supporting Information).Compared with the fluorescence signal of the total aptamers attached to rGO, the fluorescence signal of the neuroinflammation-on-a-chip administered with 1000 ng mL −1 LPS was similar, and it was observed that the fluorescence signal gradually decreased when the administered concentration was reduced to 10 ng mL −1 .It was also shown in Figure 4d that in the event of nontreated neurovascular-ona-chip, almost no fluorescence signal was visible during the experiment.Through the use of the neuroinflammation-on-a-chip sensing module, the sensing experiment yielded highly successful results in the detection and quantification of proinflammatory cytokines in a qualitative manner.
In summary, our study successfully replicated the process of neuroinflammation using a microfluidic chip system that incorporated representative brain cells and vascular endothelial cells.The results demonstrated the detrimental effects of LPS on the blood-brain barrier integrity and neural cell survival.Additionally, we developed a sensor module combining rGO and aptamers for quantitative in situ detection of proinflammatory cytokines.However, a critical step that needs to be taken is a more precise in vitro neuroinflammation model that includes microglia and monocytes through better coculture techniques. [42]To this end, we also demonstrated proof of concept regarding cell migration by injecting GFP-expressed mesenchymal stem cells into the top channel and observing their transmigration from the top to the bottom channel, where LPS was added to induce inflammation (Figure S6, Supporting Information).The neuroinflammationon-a-chip model has the potential to revolutionize our understanding of pathophysiological inflammatory responses, offering a more precise representation compared to animal models.It holds promise for advancing personalized therapeutic strategies and contributing to precision medicine by accurately mimicking the inflammatory responses observed in neurological conditions.Furthermore, our study has significant implications for multidisciplinary collaborations between engineering and medical fields.This collaboration can lead to the development of advanced technologies for drug testing, diagnostics, and the study of complex diseases in the context of a digitally enabled healthcare system (Figure 4e).Our advanced and precise technology platform enables efficient neuroinflammation modeling and accurate drug efficacy and side effect prediction.By integrating technology into healthcare, our proposed model has the potential to drive innovation and create more sophisticated platforms for studying complex diseases.

Conclusion
In this study, we have made significant progress in the development of a neuroinflammation-on-a-chip system, which has the potential to revolutionize how we investigate neuroinflammation.Our team has successfully integrated a sensor module that combines rGO and aptamers, which has allowed us to precisely and quantitatively detect proinflammatory cytokines in situ.This could be a breakthrough, as it enables us to monitor neuroinflammation with unprecedented accuracy.Our findings are exciting because they highlight the potential and effectiveness of the neuroinflammation-on-a-chip system in replicating interactions among human vascular endothelial cells, astrocytes, and neurons.By closely mimicking the complex cellular interactions that occur in the brain, this system has the potential to provide us with a much deeper understanding of the mechanisms underlying neuroinflammation.Furthermore, our study suggests that the neuroinflammation-on-a-chip system could be used in various applications, such as drug screening, brain disorder research, and personalized therapies.By offering a more accurate and efficient way to study neuroinflammation, this system could help to expedite the development of more effective therapeutic strategies.However, it is important to note that continued research is needed to improve the performance and reliability of the neuroinflammation-on-a-chip system.This will require further experimentation and optimization, ultimately contributing to a better understanding of complex brain inflammation mechanisms.Overall, our study represents an exciting step forward in neuroinflammation research.

Experimental Section
Fabrication of Microfluidic Chip: The NVU-on-a-chip microdevice used in this study was fabricated by using polydimethylsiloxane (PDMS).Silicon molds were treated with plasma and silanized with trichloro(1H,1H,2H,2Hperfluorooctyl)silane (Sigma-Aldrich, USA).Pre-PDMS with hardener (10:1 ratio) was mixed and poured on the silicon mold.After curing at 90 °C, oxygen plasma was treated (top, bottom, and porous membrane) to fabricate each piece.The dimensions of microchannels (1 mm wide × 10 mm long × 0.15 mm high) were the same in the upper and lower layers.The porous PDMS membrane contained an array of circular holes (10 μm diameter).The fabricated PDMS microfluidic device was placed onto a cover glass.Silicon tubing (Tygon 3350, Inside diameter 1/32″, Outside diameter 3/32″, Beaverton, MI, USA) with a connector (hub-free stainless steel blunt needle, 18G; Kimble Chase, Vineland, NJ, USA) was connected into each microchannel to supply cell culture medium and vacuum suction.
Culturing Neurovascular Cells on the Microfluidic Device: After oven drying for 1 h, the microfluidic device and tubing were placed on the cover glass in a UVO treatment.This was to activate the PDMS channel surface, making it hydrophilic and ready for surface coating.After treatment, the device should be fairly warm; placed under UV light until it was cooled down and ready for coating.In order to prepare the coating solution, the frozen 100 L Matrigel and Dulbecco's modified Eagle medium (with antibiotic) solution (≈100 μL) were combined into a tube.The solution was drawn into the syringe, followed by attaching the syringe filter, and finally dispensing the filtered solution.The coating solution was slowly injected into each syringe connected to both channels until no bubbles were seen inside the channels or the tubings.The microfluidic chip was incubated containing the coating solution at 37 °C for 1h.HCMEC, human cerebral microvascular endothelial cells (obtained from Sigma-Aldrich, USA) were cultured on the upper surface of the extracelluar matrix (ECM)-coated, flexible, porous membrane in the NVU-on-achip device, with the culture medium being perfused constantly (100 μL h −1 ; 0.1 dyne cm −2 ).To model the NVU, human astrocytes (Sciencell, USA) were cocultured on the opposite side of the porous ECM-coated membrane in the presence of a flowing (30 μL h −1 ) coculture medium.In addition, iPSC-NPCs were injected into the lower channel after attachment of the human astrocyte.The iPSC-NPCs were differentiated at the bottom of the lower channel with no basic fibroblast growth factor (bFGF) culture medium.To study inflammatory activation of the endothelium, LPS (15 μg mL −1 ) was introduced into the upper and lower microchannels, respectively.
Measurement of TEER Value: To measure the TEER value, two 0.1 μm platinum electrodes were prepared.One platinum electrode into the tubing was connected to the top channel and the other to the tubing was connected to the bottom channel by 30 mm.A pair of platinum electrodes were connected to the multimeter, and the resistance value was read after about 1 min.To accurately capture the average value, this work necessitated three or more recordings a day.
Permeability Test of NVU-on-a-Chip: For the permeability test, dextran-FITC (Sigma-Aldrich, USA), having a molecular weight of 3000-5000 Da was used.Specifically, dextran-FITC was injected into the top channel to each neurovascular-on-a-chip, TNF-, and LPS-treated neuroinflammation-on-a-chip chip (100 μL h −1 ).After collecting the culture medium flowing out from the top and bottom channels, the fluorescence intensity of FITC was measured with a fluorescence spectrophotometer.
Measurement of Proinflammatory Cytokines: To fabricate a sensing module in the bottom channel, rGO was immobilized.After plasma treatment with a corona treater on the bottom channel made of PDMS, rGO (1 mg mL −1 ) was put on the bottom channel (3 h, Room temperature {RT}).After 3 h, phosphatebuffered saline (PBS) was injected to remove residual rGO from the PDMS surface.The fluorescent-dye-tagged aptamers were attached onto the rGO for measuring proinflammatory cytokines.The quantity of the aptamer that reacted with proinflammatory cytokines could be measured by measuring the fluorescence sig-nal of the collected culture medium.To use this, aptamer (5 mg mL −1 ) was reacted in the bottom channel to which rGO was attached (1 h, RT).After 1 h, PBS was injected to remove residual aptamers from the rGO surface.After the neurovascular-ona-chip was completed with HCMEC, astrocyte, and neuron, the concentration of proinflammatory cytokines could be predicted by collecting the cell culture medium from the bottom channel of the chip treated with TNF- or LPS and measuring the fluorescence intensity.

Figure 1 .
Figure 1.Schematic diagram of developed neurovascular-on-a-chip. a) The schematic diagram for the necessity of NVU-on-a-chip.To study neural pathophysiology and develop potential therapeutic drugs, NVU-on-a-chip is required instead of animal models that have limitations.b) Perspective image of the developed NVU-on-a-chip.From the side/top view images, it consisted of vertically separated two parallel channels with neuronal and vascular cells which mimic in vivo neurovascular model.Neuroinflammation in the neurovascular-on-a-chip with the sensing module, which is composed of reduced graphene oxide (rGO) and fluorescent-dye-tagged aptamers.Releasing proinflammatory cytokines bound to their aptamers and detached fluorescent-dye-tagged aptamer from the GO could be measured their fluorescent intensities.

Figure 2 .
Figure 2. Characterization of the neurovascular-on-a-chip. a) Fluorescence image of the appearance of the microfluidic chip.b) Scanning electron microscope (SEM) image of the cross-sectional microfluidic chip.c) Monitoring of transendothelial electrical resistance (TEER) value for two weeks.d) Immunocytochemistry images of human cerebral microvascular endothelial cell (HCMEC) with 4',6-diamidino-2-phenylindole (DAPI) and ZO-1 staining (top row), human astrocyte with DAPI and GFAP staining (middle row), and neuron from iPSC-NPC with DAPI and Tuj-1 staining (bottom row) in the NVU-on-a-chip.The scale bars are 100 μm.

Figure 3 .
Figure 3. Induction of neuroinflammation state in NVU-on-a-chip.a) Schematic diagram of progress of neuroinflammation in NVU-on-a-chip.b) Optical and fluorescence images of healthy NVU-on-a-chip and neuroinflammation by applying LPS.c) Viability of neurons in healthy NVU-on-a-chip and neuroinflammation by applying LPS.d) Permeability test of FITC-dextran from top to bottom channel in healthy NVU-on-a-chip and neuroinflammation by TNF- and LPS.e) Schematic diagram of TEER value investigation in NVU-on-a-chip.f) Measurement of TEER value after treating TNF- and LPS (**p < 0.1, ***p < 0.001, ****p < 0.0001).

Figure 4 .
Figure 4. Integration of proinflammatory cytokine sensing module in the neurovascular-on-a-chip for in situ detection of neuroinflammation.a) Schematic diagram of sensing mechanism of proinflammatory cytokines, TNF- and IL-6.Cytokine-specific aptamers were immobilized on the rGO on the bottom layer.Secretion of proinflammatory cytokines were bound to the aptamer and they could be detected by measuring fluorescent signals.b) SEM image of GO on the bottom of the microfluidic chip.The scale bar is 1 μm.c) Raman spectrum of GO on the bottom of the microfluidic chip.The D band (1350 cm −1 ) and G band (1600 cm −1 ) are clearly displayed.d) Fluorescence intensities of collected media from bottom channels by treating LPS.Their concentrations were 1000, 100, and 10 ng mL −1 , respectively.e) Future perspective of NVU-on-a-chip.An increase in the number of key points such as microglia and monocytes is necessary for an accurate model.