Engineered human blood–brain barrier microfluidic model for vascular permeability analyses

Latest Research

The blood–brain barrier (BBB) constitutes the main access point into the brain parenchyma for endogenous and therapeutic molecules alike. Molecular transport across the BBB is highly regulated, owing principally to its unique cellular architecture with endothelial cells (ECs) lining the blood vessels, pericytes (PCs) in close contact with the ECs, and astrocytes (ACs) extending their end-feet to the abluminal side of the vessels1,2. This configuration results in the formation of tight junctions between ECs that limit paracellular transport, while transcellular transport (transcytosis) is also diminished across the BBB owing to reduced vesicular traffic2,3. At the same time, expression of efflux pumps in BBB ECs can strongly oppose small-molecule diffusion across the cell membrane4. This uniquely restrictive interface between blood and parenchyma protects the brain from toxins and pathogens, yet conversely limits the passage of therapeutics5. Models of the human BBB can, therefore, find widespread use as tools to study molecular transport in health and disease and aid the development of therapeutics with improved distribution into the brain.

Difficulties in translating results from animals to the clinic have led several research groups to focus their efforts on the design of in vitro models of the BBB. Numerous 2D EC monolayer assays using human cells in coculture (ECs and PCs or ACs) have been designed in the hope of modeling transport of therapeutic molecules across the selective endothelium6,7,8. These models, however, intrinsically fall short of recapitulating the 3D cellular organization of brain capillaries. Recently, greater emphasis has been placed on generating 3D BBB models where cells are embedded in a physiologically relevant gel matrix. While these techniques are able to generate tube-like vessel structures, the geometries and cellular organization achieved still fail to mimic the BBB in vivo9,10,11. Critically, the functional properties of these models do not fully recapitulate physiological levels, particularly in terms of protein/gene expression profiles and vascular barrier function.

We recently developed a human BBB model incorporating ECs, PCs and ACs encased in a 3D gel matrix within microfluidic devices12. The cells self-assemble into 3D vessel architectures resembling the natural BBB, with gene expression profiles and permeability values to macromolecules comparable to those observed in vivo. Methodologies developed in our group allow for quantitative assessment of permeability in the model for any molecules of interest13,14. The present protocol comprises the steps required to fabricate this BBB model and measure physiologically relevant molecular permeabilities, which may be used in academic or industry laboratories to study and predict transport across the BBB.

Development of the protocol

Recognizing the need to better recapitulate the human BBB in vitro, as well as the challenges of current 2D or 3D systems, we developed an in vitro microfluidic model of the BBB featuring perfusable microvascular networks (MVNs) composed of human ECs, PCs and ACs12. This protocol extends on our previous MVN model in which human umbilical vein ECs were used to generate microvessels in one channel of the microfluidic device, and normal human lung fibroblasts were used in the adjacent channels for stromal support15. Here, we engineer a brain-specific human microvascular model within an improved microfluidic device design that allows for culture of all BBB cell components in the same gel matrix. Induced pluripotent stem-cell-derived ECs (iPS-ECs) or primary human brain microvascular ECs (HBMECs) are cultured along with primary brain PCs and primary brain ACs to form the BBB MVNs. Several protocols have been established to obtain ECs from induced pluripotent stem cells, and endothelial state maintenance was confirmed for several passages in vitro and after implantation in vivo16,17. The successful use of iPS-ECs in the BBB MVNs may allow, in the future, the development of patient-specific models. With this aim, we validated that iPS-ECs in the BBB MVNs not only maintain their EC phenotype over time but recapitulate several key features of HBMECs in the brain microvasculature in vivo, notably perivascular cellular interactions through juxtracrine signaling, vessel morphology, gene expression profiles and functional barrier properties.

Analysis of molecular permeability is a key application of BBB models, whereby the barrier strictly controls the access of both therapeutics and biomolecules to the brain in health and disease. Here, we show that the BBB MVNs can be used to study the vascular permeabilities of model molecules and therapeutics, which are found to be of magnitudes comparable to those measured in animal models (3.1 × 10−7 cm/s for 10 kDa dextran in rat brain capillaries18 compared with 1.7 × 10−7 cm/s in the BBB MVN model, and 1.4–1.9 × 10−7 cm/s for 40 kDa dextran in rat18,19 and mouse20 brain capillaries compared with 4.2 × 10−8 cm/s in the BBB MVN model). To measure permeability in MVNs for any molecule of choice, we developed methodologies based on either imaging of fluorescence signal13, or collection and analysis of interstitial fluid14, both of which are included in this protocol. The two different methodologies may find easier application in either academic research laboratories or in industry, depending on equipment available and preferred application, making the protocol suitable for widespread use. Finally, downstream processing of the BBB MVNs included in this protocol allows for high-resolution imaging of junction and transporter proteins and cell collection for gene/protein analysis for comparison with permeability results and validation of the BBB-like properties of this platform.

The assays described in this protocol should be easily implementable in both academic and industry laboratories, as they only require standard cell culture equipment and contain several procedures that can be readily outsourced to highly accessible companies at reasonable cost, such as 3D printing companies to generate microfluidic device molds or cytokine/gene analysis companies. In addition, the small dimensions of the microfluidic devices allow for the use of reduced volumes of reagents and low cell numbers, decreasing the costs associated with each experiment and allowing for large numbers of repeats per experiment.

Applications of the method

Although this protocol describes the formation of BBB MVNs for molecular permeability analyses, the methods described are robust and flexible and can find applications in numerous other investigations. For example, the use of iPS-ECs to form BBB MVNs is a critical advantage of the model, as it allows for the formation of patient-specific MVNs. Particularly, we demonstrate that iPS-ECs adopt several key features of native BBB ECs (HBMECs) when cocultured with PCs and ACs in microfluidic devices, attesting to the relevance of this protocol in designing patient-specific BBB MVNs with potential applications in the clinic. Personalized BBB MVNs can be used to assess changes in barrier function and cellular architecture under disease conditions, such as neurological disorders where the BBB may be compromised21,22. The BBB MVNs can also be used for the permeability analysis of entities other than molecules, such as novel therapeutic carriers like nanoparticles23,24,25 or infectious agents like severe acute respiratory syndrome coronavirus 2. Similarly, the permeability analysis detailed here can be applied to other MVNs13,14 and in vitro vascular models, attesting to the flexibility of this protocol. In addition, as we have previously shown, the MVNs can be used to assess tumor cell arrest and extravasation during metastasis15,26, which can be studied in a brain-specific microenvironment using the BBB MVNs27. Inclusion of a tumor model can also be achieved in the BBB MVNs for the study of local permeability changes and therapeutic engagement28. Finally, the perfusable nature of the BBB MVNs allows for the study of expected pathophysiological flow-induced changes on endothelial function29. For example, we have shown that application of physiological levels of vascular flow through a microfluidic pump results in lower MVN permeabilities and extended model stability30.

Alternative methods

While in vivo brain investigations in animal models have offered great insight into the physiology of this organ, challenges remain in the translation of animal findings to human patients. This is evident in the high failure rates (>80%) in clinical trials of drugs previously validated to cross the BBB in animals5,21. In addition, ethical concerns and increased costs of animal experiments have driven the design of in vitro human BBB models for use as preclinical assays.

Monolayer human BBB models

The majority of in vitro BBB systems described in the literature are 2D monolayer platforms, where ECs are plated on a porous membrane or on top of a thin hydrogel layer (often collagen) and, in some cases, with a combination of PCs, ACs or neurons8,31,32,33,34. The simplicity and ease of use of these systems have increased their popularity and allow for the design and implementation of experiments that require few cells and reagents and yield results in short timeframes7,35. Permeability can be easily measured by direct analysis of molecular concentrations in fluid. Thus, this type of assay has proved to be amenable for high-throughput drug testing in a simple platform or the study of cancer cell transmigration in 2D36. However, the capability of these systems to recapitulate the brain microenvironment to generate a clinically relevant model remains limited. Particularly, 2D EC monolayer models fail to mimic key cellular interactions present in the brain between ECs, PCs and ACs. In addition, the morphological features of brain capillaries pertaining to their geometry and structural organization are lacking in 2D models, resulting in shortcomings in their functional properties. This is notably observed in measurements of permeability, which are often one to three orders of magnitude higher in 2D BBB models compared with results in vivo, despite the presence of confluent endothelial monolayers13.

Tube-like human BBB models

To address the limitations of 2D monolayer models in their ability to mimic the BBB in both architecture and function, several groups have shifted their efforts to the design of tube-like 3D vessel structures. These models are either generated from hollow structures in 3D hydrogel produced with a needle pull-out method9,37,38 or within hollow channels in microfluidic systems10,11,39. ECs are then perfused to line the internal walls of the channels and assemble into monolayers that adopt a tube-like structure. In these systems, PCs, ACs or neurons can be perfused prior to the ECs to ensure that the cells are positioned on the abluminal side of the tube-like vessel9, or can even be directly embedded in the hydrogel8,10. As such, tube-like models offer an additional layer of complexity and accuracy when compared with 2D monolayer models. These models also allow for quantification of vascular permeability via confocal microscopy of fluorescence signal9,10, and the effects of microenvironmental changes on vessel stability and barrier function can also be assessed through measurements of impedance-based permeability39. Nevertheless, the tube-like structures produced lack physiological microvessel diameter sizes, as they are typically as large as 600–800 µm in diameter9,39, so that the local surface curvature at the scale of a cell is not far from that of a planar monolayer. In addition, although vascular permeabilities in these structures are shown to decrease with the presence of PCs or ACs9, their values are still relatively close to those obtained in 2D models and one to two orders of magnitude higher than typical permeabilities in vivo3,9,10. While monolayer and tube-like vessels are highly appropriate tools for high-throughput quantification of barrier permeability in a simple setup, the BBB MVN model presented here features physiologically relevant vessel morphology in 3D, appropriate BBB gene expression and low values of permeability. These characteristics make it clinically relevant and highly amenable for the study of receptor-mediated transport, cell transmigration in 3D and interactions of stromal cells with the endothelium12,27.

Alternative measurements of permeability

Vascular permeability has been measured either by assessment of electrical impedance of EC monolayers or by monitoring of molecular concentration in the abluminal interstitial space over time40. The first method, which measures transendothelial electrical resistance (TEER), requires the ECs to be plated in contact with electrodes, after which measurement can be performed with great ease over time. For this reason, the method has found some application in monolayer BBB models3,31,39. One key advantage to TEER lies in its ability to provide real-time monitoring of barrier permeability with a simple electrode setup41, a temporal resolution often difficult to achieve with other methods. TEER is also used as a proxy for EC junctional integrity, providing a general assessment of paracellular transport (i.e., in between junctions) across the endothelium, with no regard to molecular specificity or transcellular transport (i.e., transcytosis of larger molecules and passive membrane diffusion of small molecules). Measurements of permeability through molecular flux, on the other hand, are sensitive to those factors and can be performed in both 2D and 3D BBB models, unlike TEER, which is limited in 3D because of difficulties in electrode positioning42. However, physiologically low paracellular permeabilities are necessary to not overshadow transcellular transport and allow for its quantification. Using MVN models, we were able to show improved physiological vascular permeabilities compared with 2D and 3D tube-like vessel platforms, as well as the ability to finely assess transcellular transport and separate it from paracellular transport13,14.

Advantages and limitations of the protocol

The BBB MVN model described in this protocol and the associated methods to evaluate vascular permeability hold several advantageous characteristics:

• ECs, PCs and ACs are used to generate 3D triculture microvessels with small diameters (~10–40 µm) and appropriate BBB gene expression profiles, thus offering greater physiological relevance when compared with conventional EC monolayer or tube models

• Human cells (stem-cell-derived or brain primary cells) can be exclusively used in this model, increasing the relevance of this platform to predict molecular permeability in the human brain in health and disease compared with animal models

• Analysis of permeability in the BBB MVN models is simple to perform and highly quantitative compared with analysis in animal models, where difficulties in imaging through thick tissues and animal drift greatly complicate the measurements

• BBB MVN permeability can be measured with high spatiotemporal control to assess changes over time due to specific treatments or challenge with disease-associated factors like proinflammatory cytokines13,43

• Transmural flow can be applied in the BBB MVNs to assess molecular permeability under physiological intravascular pressure conditions44

• Physiologically low paracellular transport in MVNs allows for assessment of transcytosis. Specific transporter or efflux proteins can be inhibited to study their role in molecular transport across the endothelium13

• Clearance of perfused molecules by ECs can be assessed through colocalization analysis in different cell compartments, such as the early endosome and lysosome13

• The technical expertise to form the BBB MVNs is low compared with other techniques for in vitro brain vascular models that require 3D cell printing techniques, sacrificial gels or needle pull-out techniques

• Cell isolation for quantification of gene and protein expression levels is simple compared with animal models, which require specialized knowledge of invasive techniques to isolate cells from the brain and additional expensive equipment

• The BBB MVN model can be engineered using standard cell culture equipment and is inexpensive with low reagent volumes and cell counts needed per device (considered a biological repeat). Therefore, information on BBB MVN properties can be obtained from numerous repeats at greatly reduced costs (<0.40 USD per device for fabrication costs15 and <1.50 USD per device for device seeding and cell culture costs) compared with animal models, which are at least 50 times more expensive (>25.00 USD per animal for purchase and >80.00 USD per animal per month for housing45). In addition, users can also opt for commercially available devices (e.g., 3D cell culture chips by AIM Biotech) to generate the BBB MVNs following the protocol below

A key limitation to the BBB MVN model, as in any other in vitro model, is the only partial recapitulation of the BBB microenvironment in vivo. The model fails to fully capture the interactions of perfused molecules with other host cells in the brain circulation and parenchyma; for example, immune cells that play a role in the clearance of therapeutics and other molecules46. The flexibility of the BBB MVNs allows for modifications of the model to include additional cell types (e.g., immune cells in the circulation or neurons15,47). Another key limitation of the model is that perfusion of molecules in the BBB MVNs for long-term studies is not feasible due to a lack of physiological clearing mechanisms in the gel matrix (i.e., glymphatic flow in the BBB in vivo48) and diffusion of the perfusate across the side monolayer. As such, only short-term studies of permeability can be performed in the BBB MVNs. Nevertheless, short-term permeability analysis remains highly effective to rank transport of different therapeutics across the BBB MVN endothelium. Changes in permeability over long time periods (days) may still be assessed by washing out the perfusate with fresh medium and attempting permeability analysis with the same perfusate on another occasion.

Experimental design

Characterizing molecular transport across the BBB requires a brain microvascular model featuring morphological and functional properties comparable to those of brain capillaries in vivo, the capability for high spatiotemporal resolution imaging, and extraction of the cells contained in the 3D gel matrix for protein/gene analyses and comparison with ex vivo tissue and cells. To this end, we have developed an in vitro BBB microfluidic model with 3D MVNs composed of ECs (iPS-ECs or HBMECs), PCs and ACs that presents those features and capabilities. The protocol consists of four major modules: (i) microfluidic device fabrication and BBB MVN formation (Steps 1–20), (ii) measurement of vascular permeability (Steps 21–46), (iii) imaging of proteins and cell architecture (Steps 47–60), and (iv) cell collection for gene/protein analyses (Steps 60–67).

Device fabrication and BBB MVN formation (Steps 1–20)

Two different device types, both featuring a central gel channel and two adjacent medium channels, are employed in this protocol: (i) the microdevice and (ii) the macrodevice (Fig. 1, Supplementary Figs. 1 and 2, Supplementary Table 1 and Supplementary Data 1 and 2). Central channels in microdevices are delineated by upright microposts separated by 200 µm to allow for surface-tension-assisted filling of cell-laden fibrin gels. In the macrodevices, the additional height of the device makes it possible to use a partial wall from the top surface of the device chamber to separate the central and side channels, allowing for an uninterrupted gel–medium interface.

The choice of device for use in the formation of BBB MVNs should be dictated by the user’s end application. All methodologies described in this protocol can be applied for both device types, although we point out the suggested device for each application throughout. Generally, permeability measurements are easier and less variable when performed in the macrodevice, due to the increased width of the central gel channel limiting diffusion of perfusates from the side channels over time. The macrodevice also contains more cells for gene/protein analysis, and the increased size of the gel makes it easier to collect for further histological sectioning. However, the increased size of the macrodevice also makes it difficult to perform effective immunofluorescence staining—the microdevice should be used, instead, for high-resolution imaging of immunofluorescence-labeled proteins and cell interactions. The microdevice also utilizes less than half of the cell numbers required for the macrodevice, resulting in up to three times more repeats for the same starting cell numbers. The number of devices seeded per experiment is at the discretion of the user, but we have concluded that 100 independent microdevices or 35 independent macrodevices can be seeded in one sitting without compromising cell viability.

We have validated that both human iPS-ECs and primary HBMECs can be employed to form the BBB MVN model, and have compared the protein and gene expression profiles of both ECs in the microfluidic chips after 7 d of culture. Human iPS-ECs or HBMECs, along with human primary brain PCs and ACs suspended in fibrin gels, are injected into the central channel of the device (Fig. 1). During device culture, iPS-ECs or HBMECs elongate, connect with neighboring cells, and form lumen structures that stabilize within 7 d. Simultaneously, PCs and ACs adopt a perivascular position along the newly formed blood vessels as observed in vivo, with PCs wrapping around the lumens and ACs extending their end-feet to touch the abluminal side of the vessels (Fig. 1). The adopted 3D cellular architecture and resulting vessel morphologies recapitulate the properties of BBB capillaries in vivo, and maintain stable dimensions starting at day 6 of culture (Extended Data Fig. 1). The addition of a monolayer of ECs at the medium–gel channel interface on day 4 helps prevent diffusion of solutes through the gel surrounding the MVNs when measuring permeability (Fig. 1).

Permeability analysis (Steps 21–46)

A key advantage of the BBB MVNs12 and other, tissue nonspecific MVNs developed in our group13,14 is the capability to recapitulate vascular permeabilities comparable to those measured in animal models. Formation of physiological tight junctions and the expression of a functional glycocalyx and basement membrane limit paracellular transport between ECs13, which is key to the low permeability values measured in our systems for macromolecules compared with simpler 2D EC monolayer models (5.6 × 10−7 cm/s for 10 kDa dextran in 2D HBMEC monolayer models49 compared with 1.7 × 10−7 cm/s in the BBB MVN model, and 2.0–9.0 × 10−7 cm/s for 40 kDa dextran in 2D HBMEC49,50 or iPS-EC monolayer models8 compared with 4.2 × 10−8 cm/s in the BBB MVN model). Expression of BBB-specific transporters and efflux proteins makes the BBB MVN model also amenable to the study of small-molecule transport by passive diffusion across the cell membrane and large-molecule transcytosis. Permeability in the BBB MVNs may be measured in one of two ways, described below with their relevant experimental parameters, or a combination of the two for additional validation of the results (Supplementary Method, Supplementary Table 2 and Supplementary Software 1 and 2).

Permeability analysis—confocal microscopy of fluorescence signal

Imaging through confocal microscopy can be used to measure permeability of molecules conjugated to fluorophores (Fig. 2), under the core assumption that fluorescence intensity varies linearly with molecular concentration13. Numerous molecules are commercially available as preconjugated to fluorophores, such as the ones used in this protocol, making knowledge of fluorescence labeling techniques unnecessary. The molecule of choice is dissolved in cell culture medium and perfused through the BBB MVNs, after which the sample is imaged in the same locations at different timepoints. Image analysis is used to yield the average fluorescence intensity in the vasculature and matrix over time, as well as relevant morphological parameters. The permeability, P, is calculated as51:

$$P = frac{1}{{{{{mathrm{{Delta}}}}}t}}frac{{V_{{{mathrm{m}}}}}}{{SA_{{{mathrm{v}}}}}}frac{{{Delta}I_{{{mathrm{m}}}}}}{{{Delta}I}}$$

(1)

where Δt is the time between timepoints, ΔIm = Im,2Im,1 is the increase in average fluorescence intensity in the gel matrix with volume Vm, SAv is the surface area of the vasculature, and ΔI = Iv,1Im,1 is the difference in average intensity between the vasculature and matrix at the start of the measurement. A confocal microscope is strictly required for this methodology owing to the 3D morphology of the BBB MVNs and resulting 3D analysis. Equation 1 assumes that the matrix is homogeneous and permissible, as previously confirmed14. Bleaching of the fluorescent solutes used in this protocol has rarely been observed, likely due to the short imaging times involved in the methodology described, yet it may modify the permeability values measured. Assuming a constant solute concentration in the vascular space through the duration of the measurement, photobleaching may be observed as a decrease in the vascular fluorescence intensity, and Eq. 1 can be applied by replacing Im,2 with Im,2 * = Im,2 × (Iv,1/Iv,2), where Iv,2 is the vascular fluorescence intensity at the second timepoint. A further assumption is that no transport occurs across the boundaries of the imaged region. This is supported by the fact that the vessel density and fluorescent dye concentration are relatively homogeneously distributed. Variations in either variable could give rise to errors in the measurement.

As with any fluorescence image analysis technique, image acquisition on the confocal microscope must be set up so as to not oversaturate the signal, which would result in loss of information. Even when collecting signal just below the saturation limit, sufficient time is required between timepoints to measure a substantial increase in fluorescence intensity compared with noise in the matrix (Extended Data Fig. 2), due to the low permeabilities typically measured in MVNs. Conversely, prolonged time between timepoints will negatively impact the measurement owing to diffusion of the perfusate from the side channels through the monolayer. An optimal time window, therefore, exists for the measurements depending on the solute molecular weight, reported in Table 1 for a range of molecules in the macrodevice. In the microdevice or commercially available AIM Biotech chips, due to the smaller gel channel, a substantial increase in matrix fluorescence intensity can be observed sooner due to diffusion from the side channels, and the lower bound of the time windows in Table 1 should be used to minimize corruption of the measurements and abnormally high permeabilities resulting from compound effect of permeability through the BBB MVNs and diffusion through the monolayer (Extended Data Fig. 2).

The size of the imaged sample volume can also affect the quality of the permeability measurements by confocal microscopy. Scattering of the fluorescence signal at increasing depths lowers the intensity measured, resulting in a reduced signal-to-noise ratio (Extended Data Fig. 3). Irrespective of the height of the device, therefore, the imaged volume used for the analysis should be thicker than 50 µm only for especially bright conjugated fluorophores. In the xy-plane, the area imaged should be as large as possible as to best capture the average morphology of the BBB MVNs. For this protocol, we used a 10× confocal objective (approximate field of view 1,250 × 1,250 µm), although higher-resolution objectives may be used, provided they allow for imaging of areas larger than 600 × 600 µm, for which consistent results are expected (Extended Data Fig. 3). Image resolutions as low as 640 × 640 pixels are adequate to effectively capture boundaries between BBB MVNs and gel matrix. These parameters and considerations are provided in Table 2 for user reference during experimental design.

Permeability analysis—interstitial fluid collection

The use of confocal microscopy may not be available or preferred in some laboratories, which may instead be more accustomed to analysis of perfusates in fluid through techniques such as ELISA and mass spectrometry. In addition, conjugation of molecules to fluorophores may alter their physicochemical properties and result in different permeabilities across the vascular endothelium52. Therefore, we have developed a methodology that can be applied in the MVNs to collect interstitial fluid for direct analysis of labeled and unlabeled molecules14. The BBB MVNs are pressurized to a physiological net pressure across the endothelium (Fig. 3). This pressurization in turn produces physiological transmural flow across the vascular endothelium of the BBB MVNs, whereby fluid flow continues across the gel matrix and exits the device through the gel ports, where it can be directly collected.

The perfusate concentration in the interstitial fluid collected, c, is a fraction of the concentration perfused in the BBB MVNs, c0. This methodology assumes that no solute will become bound to the matrix, so that the difference between c0 and c is entirely due to solute partition across the vascular wall. This assumption is supported by the large permissivity of the fibrin gel matrix, as previously described14. The ratio between c0 and c can be used to compare different molecules, as it is often done when testing molecular distribution in preclinical animal models53. In the BBB MVNs, the ratio can also be used to yield the BBB MVN permeability, P, through the relationship:

$$frac{c}{{c_0}} = frac{{P + L_{{{mathrm{p}}}}{{{mathrm{{Delta}}}}}p(1 – sigma )}}{{P + L_{{{mathrm{p}}}}{{{mathrm{{Delta}}}}}p}}$$

(2)

Here Δp is the pressure applied, Lp is the hydraulic conductivity of the BBB MVNs (Extended Data Fig. 4) and σ is the reflection coefficient of the perfusate. The values that can be assumed for this analysis are reported in Table 3.

Similar to the methodology using confocal microscopy, permeability analysis through pressurization of the BBB MVNs has an optimal window for data collection. At short times (1–3 min) after pressurization, the fluid collected is the one originally present in the matrix, where the perfusate concentration is virtually zero. At long times (typically >15 min), diffusion and convection of the perfusate through the side monolayer will corrupt the measurements and yield abnormally high concentrations. Collection of interstitial fluid every 1 min after pressurization for 10 min will ensure that consistent values of perfusate concentration can be measured (Fig. 3).

The suggested applied pressure of 1 kPa may also be changed by the user depending on the available pressure regulators. However, a higher pressure will result in a shortening of the optimal window for the measurements, and pressures >2.5 kPa have not been attempted by the authors and may result in damage to the BBB MVNs.

Pressurization of the BBB MVNs can be combined with fluorescence imaging by confocal microscopy to measure effective permeability values under physiological transmural flow conditions. We have shown that the effective permeability in these conditions is higher for small molecules that can squeeze through EC junctions and be transported by flow, while it does not change for large molecules like plasma proteins and monoclonal antibodies14.

Imaging of proteins and cellular architecture in whole or sectioned gel (Steps 47–60)

Visualization of proteins associated with transport regulation through the BBB MVNs can be performed to assess their presence and localization. Proteins can be visualized in the models in 3D by fixing and processing devices for immunofluorescence staining, in a similar manner to tissue staining. The small dimensions of the devices result in substantial reductions in reagent volumes needed to perform immunofluorescence staining and, therefore, costs of the method. Additionally, the presence of two adjacent medium channels flanking the gel region allows for the application of a hydrostatic pressure drop across the central gel channel, ensuring appropriate perfusion of antibodies and optimal immunostaining conditions. Another advantage of this method when using the microdevice lies in the capability for high spatial resolution imaging of the entire gel thickness with BBB MVNs (Fig. 1). The ability to visualize these proteins is extremely valuable to identify the 3D organization of the BBB MVN cells and specific cellular interactions that can result in improved BBB-like protein expression1,54.

The limitations of confocal microscopy imaging, in terms of scattering of the fluorescence signal at large depths in the devices and resolution thresholds, can be overcome in the BBB MVNs by extracting the gel from the devices for histological sectioning. This approach, similar to ex vivo tissue sectioning, allows for high-resolution (60×) confocal imaging of MVNs at any depth13, or for alternative modes of imaging to be applied, such as histology (e.g., hematoxylin and eosin staining13,14) and electron microscopy for nanometer-resolution imaging13. These methods can be used to differently evaluate 3D cell architecture in the BBB MVNs and image subcellular transport processes.

Cell collection for gene and protein analyses (Steps 60–67)

Gene and protein expression profiles of the BBB MVN cell components can be measured for comparison with primary cells or to assess changes due to various treatments. Here, extraction and isolation of iPS-ECs was of interest to quantify gene expressions via real-time quantitative reverse transcription polymerase chain reaction (qRT-PCR) of relevant junctional and transporter markers characteristic of BBB ECs2,5,55. Cells can be extracted from the 3D gel at any timepoint during device culture to evaluate gene/protein levels; however, we have previously shown that MVN maturity is achieved at day 7 of culture12,13. Pooling the gels of three macrodevices per biological repeat ensures that enough ECs can be isolated via fluorescence-activated cell sorting (FACS) to obtain suitable RNA levels necessary for appropriate gene quantification via PCR (Fig. 1). The capability for EC isolation following 3D culture with PCs and ACs can provide unique insight into the functional properties of brain microvessels.