A Hybrid Nanofiber/Paper Cell Culture Platform for Building a 3D Blood-brain Barrier Model

Abstract

The blood brain barrier (BBB) protects the central nervous system from toxins and pathogens in the blood by regulating permeation of molecules through the barrier interface. In vitro BBB models described to date reproduce some aspects of BBB functionality, but also suffer from incomplete phenotypic expression of brain endothelial traits, difficulty in reproducibility and fabrication, or overall cost. To address these limitations, we describe a three-dimensional (3D) BBB model based on a hybrid paper/nanofiber scaffold. The cell culture platform utilizes lens paper as a framework to accommodate 3D culture of astrocytes. An electrospun nanofiber layer is coated onto one face of the paper to mimic the basement membrane and support growth of an organized two-dimensional layer of endothelial cells (ECs). Human induced pluripotent stem cell-derived ECs and astrocytes are co-cultured to develop a human BBB model. Morphological and spatial organization of model are validated with confocal microscopy. Measurements of transendothelial resistance and permeability demonstrate the BBB model develops a high-quality barrier and responds to hyperosmolar treatments. RNA-sequencing shows introduction of astrocytes both regulates EC tight junction proteins and improves endothelial phenotypes related to vasculogenesis. This model shows promise as a model platform for future in vitro studies of the BBB.

TOC Caption

Huang and coworkers describe a cell culture platform comprised of paper coated with nanofibers. Culture of a two-dimensional layer of endothelial cells atop the nanofiber layer and a three-dimensional layer of astrocytes within the paper fibers creates a model for studying the blood-brain barrier. Characterization via microscopy, resistance and permeability measurements, and differential gene expression underscore the promise of this model for in vitro studies of the blood-brain interface.

Graphical Abstract

1. Introduction

The blood-brain barrier (BBB) is a low-permeability interface between the central nervous system (CNS) and circulating blood. Necessary nutrients and hormones are allowed to flow between the CNS and blood, while the toxic compounds and pathogens are regulated.^{[1, 2]} Dysfunction of the BBB can lead to ion dysregulation, altered signaling homeostasis, and neuron dysfunction/degeneration;^[3] this implicates the BBB as a critical aspect of neurological diseases including Alzheimer’s^[4] and Parkinson’s, ^[5] as well as brain tumors.^[6] The protective nature of the BBB is unfavorable for drug delivery to the CNS, which limits neurotherapeutic applications.^{[7, 8]} Tissue models derived from cell culture have been developed to better understand the structure and function of the BBB, to optimize drug delivery strategies, and to aid in the discovery of effective treatments for neurological diseases.^{[9, 10]}

The BBB consists of a complex spatial arrangement of cell types, with brain endothelial cells lining capillaries, along with pericytes and astrocytes present in the basement membrane.^{[11, 12]} Endothelial cells control BBB barrier properties through formation of tight junction contacts between cells. Endothelial cells are partially surrounded by pericytes, and both are further covered by the basement membrane. The end-feet of astrocytes wrap around the brain capillaries and play an important role in regulating the BBB function and neuronal communication.^[13]

A comprehensive understanding of the BBB in vivo has been elusive, hampered by difficulty in decoupling the complexity of cellular interactions. In vitro models provide a promising route to study the BBB and have typically been developed by co-culturing BBB component cells on porous transwell membrane supports for studying BBB characteristics.^{[10, 14, 15]} To date, in vitro BBB models have displayed only modest functionality and relevance to disease processes due to lack of cell-cell or cell-matrix interactions. To better mimic BBB complexity, various models that make use of 3D cell culture and microfluidic systems have been developed.^[16–21] These models have shown promise for replicating the BBB, especially the components of microvascular structure. However, many approaches require delicate and complex substrates that are laborious to design and fabricate. Further, due to specialized substrate design, reproducibility of these models and lack of compatibility with other characterization techniques prove limited relative to traditional transwell models. For these reasons, a 3D BBB model that is easy and inexpensive to fabricate, provides a better representation of in vivo conditions, and is compatible with various cell-based assays is needed for further studies of barrier properties and function.

Paper is an emerging cell culture substrate with benefits that include biocompatibility, cost-efficiency, and ease of large-scale manufacturing.^{[22, 23]} Paper provides a fibrillar framework with a scale suitable to promote 3D cell culture.^[23–26] Paper-based cell culture models have found wide applications that include high-throughput drug screening,^[27–29] disease models,^{[30, 31]} and cell invasion studies;^[32–34] however, the relatively large pore diameter and porosity of papers limits utility in formation of confluent endothelial cell monolayers needed in a BBB model. Electrospinning of natural and synthetic polymers has been used to create 3D fibrous scaffolds with tunable fiber diameters (scales of ~10 to ~1000 nm) and densities for various cell models and tissue engineering.^[35–42] The combination of these 2 scaffold types, paper and electrospun nanofibers, affords the possibility of a confluent endothelial cell layer (atop nanofibers), with an underlying porous framework (paper) that can promote 3D growth of astrocytes and other neuronal cells.

In this study, we demonstrate a hybrid paper/nanofiber-based 3D cell culture platform for building a BBB model. The platform consists of a commercial high-porosity paper as the scaffold that supports an electrospun nanofiber layer on one side. Coating the paper and nanofiber with Matrigel forms a suitable mimic for the basement membrane.^[43–45] Human induced pluripotent stem cells (hiPSCs)-derived endothelial cells (iPS-ECs) and astrocytes (iPS-Astros) were co-cultured in the platform to develop a human BBB model. Immunofluorescence imaging, resistance/permeability measurements, and gene expression analysis emphasize the importance of the 3D nature of the platform. This model presents unique and simple approach to build in vitro 3D models that recapitulate important aspects of the BBB.

2. Results and Discussion

Figure 1 shows a cartoon (Figure 1a) and the procedure (Figure 1b) for development of the 3D BBB model described here. The hybrid paper/nanofiber cell culture platform consists of a sheet of paper with a layer of electrospun polystyrene (PS) nanofiber deposited onto one face of the paper. In the natural BBB basement membrane, specific extracellular matrix (ECM) proteins such as laminin, collagen IV, and proteoglycan form thin mesh-like structures cross-linked to randomly oriented nanoscale fibrils.^[46] Here, we use Matrigel, a gelatinous mixture of the same ECM proteins,^[47] to coat the PS nanofibers and create a structural and chemical surrogate for the basement membrane.^{[43, 48]} The paper sheet under the nanofiber layer is utilized as a scaffold to support growth and development of 3D astrocytes, also with the aid of Matrigel, as hydrogels of ECM have been shown to improve cell-substrate interactions and promote development of 3D structures that more closely mimic in vivo phenotypes.^[47] The nanoscale, planar morphology of the PS nanofiber layer supports growth of a confluent, two-dimensional (2D) layer of endothelial cells. Because the nanofiber layer is also thin and porous, interaction between endothelial cells and astrocytes, which are cultured in the microfiber space of the paper, are promoted, providing an environment that more closely mimics the spatial relationship of the in vivo BBB relative to commonly used polymeric membranes (vide infra). The culture platform can be adapted to transwell supports of any size and the 2D endothelial monolayer design ensures the compatibility of this model with techniques and cell-based assays for evaluating traditional in vitro BBB models, and studies of drug delivery and disease.

Figure 1.

(a) Schematic of the 3D paper/nanofiber-based BBB model. (b) Schematic diagram for derivation of iPS-ECs and iPS-Astros from hiPSC sources and building the BBB model from the two cell types.

2.1. Electrospinning PS Nanofibers on the Paper

The first step of fabricating the paper/nanofiber hybrid cell culture platform is electrospinning a PS nanofiber layer onto one side of a paper sheet. In this study, 40 μm-thick Whatman grade 105 lens paper with high porosity was used as the paper scaffold. A nanofiber mat with uniform fiber widths and pore sizes is necessary to support growth of the endothelial cell layer. The diameter of electrospun fibers is strongly correlated to solution concentration, which presents a challenge for nanoscale fiber formation. When polymer concentrations are reduced significantly, hybrid electrohydrodynamics between electrospinning and electrospraying can be observed, leading to possible decreased entanglement of the polymer chains and formation of beaded fibers.^{[49, 50]} Indeed, a standard setup for electrospinning nanoscale PS, with PS concentrations less than 15% (w/v) and flow rates less than 15 μl/min (from a syringe pump), led to irregular fiber sizes and significant beading (Figure S1a). Formation of suitable PS nanofibers was realized by pressure driven fluid flow through a silica capillary (150 μm inner diameter) with a platinum wire inserted in solution for electrical contact (Figure S2). PS solution concentrations of 10% were used, with 1% tetrabutylammonium bromide (TBAB) electrolyte added to stabilize the Taylor cone and reduce beading (see Figure S1b and c). In addition, the electrospinning process was done in a humidity-controlled box and all spinning was done in less than 20% humidity. Finally, the resulting nanofiber sheets were large compared to the size needed to fabricate the transwell, so it was possible to spray multiple sheets from the same solution in one day. These factors and the strict control of solution concentration (10% PS with 1% TBAB) helped to reduce the variation between batches of the nanofibers used in these studies.

2.2. Characterization of Paper/Nanofiber Cell Culture Platform

Figure S3b and S3c shows scanning electron micrographs (SEM) of the top view and side view of the paper sheet with electrospun PS nanofiber, respectively. The paper contains pores with sizes larger than 100 μm and fiber widths in the range of 5–10 μm. The lens paper used here has 40 μm thickness and 80% reported void volume,^[26] providing sufficient room for 3D growth of astrocytes.^{[51, 52]} The PS nanofiber layer contains pore sizes less than 5 μm, which is smaller than both endothelial cells and astrocytes and suitable to create a layered structure that prevents cell invasion between the layers. The pore size is also not too small to block the interaction between endothelial cells and astrocytes for proper development of barrier function. The PS-nanofiber side of the culture platform is defined as the apical side, and the paper only side is defined as the basal side.

The Matrigel coating of the nanofiber layer was performed by evenly spreading growth factor-free Matrigel on top of the nanofiber layer at 200 μg/cm², followed by air-drying. Since the nanofiber layer has higher surface area and fiber density compared to the lens paper, Matrigel accumulates at the nanofiber layer during the drying process, as demonstrated in Figure 2a. On the apical side (Figure 2b), the Matrigel coating forms a continuous film, providing a surface favorable for growth of a 2D endothelial monolayer. Electron microscopy of the basal side (Figure 2c) indicates the fibrous 3D nature of the paper is retained after the Matrigel coating.

Figure 2.

(a-d) Scanning electron micrograph of (a) the whole paper/nanofiber-based cell culture platform; (b) the apical side of the platform with the Matrigel-coated nanofiber layer for endothelial cell culture; (c) the basal side of the platform with the network of paper fibers creating space for 3D astrocytes culture; (d) closer view of the Matrigel-coated nanofiber layer. (e) SICM topography image of the Matrigel-coated nanofiber layer in an aqueous electrolyte.

Figure 2d shows a closer view of the Matrigel-coated nanofiber layer. The average width of the nanofibers was 210 ± 24 nm (n = 50) with the pore size of nanofiber layer 3 ± 1 μm (n = 50), and the thickness of the layer was 2.8 ± 0.3 μm (n = 5) from the batches across different days with an electrospinning spray time of 60 min. In addition, by changing the duration of electrospinning, the thickness of the nanofiber layer can be accurately controlled, ranging from 0.2 μm to 5.5 μm (Figure S4). This unique feature makes the paper/nanofiber-based culture platform a promising substrate to study the effect of the basement membrane on the interaction between two different cells, with the caveat of a trade-off between the thickness and surface roughness of the nanofiber layer. If the nanofiber layer is too thin, an uneven surface, attributed to paper fibers protruding into the nanofiber layer, may result in adverse effects on the development of cell monolayers (Figure S5). As more fibers were electrospun onto the surface, the surface roughness improved. Under conditions employed here, the nanofiber layer used for building the BBB model was made from 60 min of electrospinning (2.8 ± 0.3 μm thickness, n = 5). Although the layer was thicker than a typical in vivo basement membrane (~0.1–1 μm),^{[46, 53]} the nanofiber-Matrigel layer developed here is a significant advance over most electrospun fiber substrates with thicknesses of tens of microns.^{[43, 54]}

Surface properties of the nanofiber layer in aqueous electrolyte were further characterized by topographical mapping with scanning ion conductance microscopy (SICM). Figure 2e and Figure S6 demonstrate SICM topographical images and the corresponding 3D maps of the nanofiber layers. Importantly, structure and morphology of the hydrated nanofibers is maintained, and agrees well with structures observed in vacuo. The maximum height variation of the surface was ca. 6 μm for a 40 μm x 40 μm scan area (Figure S6), indicating an acceptable roughness for the growth of endothelial cells.^[55]

2.3. Characterization of iPS-ECs and iPS-Astros

Figure 1b illustrates the timeline and conditions used to generate iPS-ECs and iPS-Astros from hiPSCs. Human iPS-ECs and iPS-Astros were chosen for the studies here, as the complexity of the human brain and species differences in brain uptake suggests human-derived in vitro models are more useful for drug and disease studies than animal models.^{[56, 57]} Further, hiPSCs provide promising sources for generating low-cost and robust human cell models.^{[56, 58]} To verify successful differentiation of hiPSCs into iPS-ECs and iPS-Astros, immunofluorescence analysis was applied to analyze the presence of endothelial and astrocyte markers in these two cell types. Figures 3a and 3b show confocal images of iPS-ECs and iPS-Astros cultured on a Matrigel-coated surface with 4′,6-diamidino-2-phenylindole (DAPI) used to label the positions of cell nuclei. The use of fluorescently labeled antibodies for platelet endothelial cell adhesion molecule (PECAM-1) identified iPS-ECs as positive for the endothelial marker as well as formation of a confluent monolayer with no obvious intercellular spaces and a morphology representative of endothelial cells. Fluorescently labeled antibodies to glial fibrillary acidic protein (GFAP) were used to identify iPS-Astros, which showed a strong signal for the astrocyte marker and displayed the typical star shaped morphology of glial cells, with end-feet features clearly developed. Biocompatibility of the cell culture platform with iPS-ECs and iPS-Astros was tested with calcein AM cell viability assays, where iPS-ECs and iPS-Astros were separately cultured on the apical side and basal side of the platform, respectively. As shown in Figure S7, both iPS-ECs and iPS-Astros displayed high cell viability on the platform, with morphologies that appear similar to that observed in Figure 3a and 3b.

Figure 3.

(a, b) hiPSC-derived iPS-ECs (a) and iPS-Astros (b) used in this study, cultured on a Matrigel-coated surface. (c, d) Morphology of the iPS-Astros cultured (c) on a traditional Matrigel-coated porous PET membrane and (d) inside the paper of the cell culture platform. Scale bars: 100 μm.

To demonstrate the influence of the 3D paper scaffold on astrocyte morphology, iPS-Astros were cultured on two substrates mounted on the same transwell for comparison: (i) a Matrigel-coated porous polyester (PET) membrane used to build the traditional in vitro BBB models and (ii) the culture platform developed here (also with Matrigel coating as described above). Figures 3c and 3d show the resulting morphology of iPS-Astros using the GFAP label. A 2D monolayer of iPS-Astros is formed on the PET membrane. Despite slightly lower confluency, iPS-Astros cultured on paper developed more 3D morphologies with many cells appearing to overlap, indicating cells grew beyond a 2D monolayer. Qualitatively, the morphology of iPS-Astros on the PET membrane appears less star-shaped with less development of end-feet structures, as compared to the cells grown in the paper. These results further demonstrate the paper/nanofiber-based culture platform as a suitable substrate for generating the 3D BBB model.

2.4. Building the BBB model

The paper/nanofiber-based BBB model was generated by co-culturing iPS-ECs and iPS-Astros in the culture platform mounted on a customized transwell setup. A Matrigel-coated nanofiber layer was treated with a mixture of collagen IV and fibronectin beforehand, a common method used to purify hiPSCs-derived endothelial cells^{[59, 60]} and brain endothelial cells in primary culture.^[61] On the basal side, iPS-Astros were first seeded onto the platform and allowed 4–6 hrs to fully adhere to the paper scaffold. The transwell setup was then inverted and iPS-ECs were seeded on the nanofiber layer. After 7-days of co-culture (Figure 4 and Figure S8, replicant experiments performed on different days), iPS-ECs and iPS-Astros were positive for PECAM-1 and GFAP, respectively, with elongated morphologies observed in iPS-Astros. Figure 4d shows a cross-section of the BBB model, 3D-rendered from stacks of confocal images recorded in the z-direction. A flat 2D monolayer is observed for the iPS-ECs, and a 3D multilayer structure of iPS-Astros is observed in the paper scaffold below the iPS-EC monolayer. There were no signs of iPS-ECs penetrating the nanofiber layer and entering the astrocyte spaces, though there were minimal observation of fluorescence signals indicating iPS-Astros reached the iPS-ECs layer, probably because the width of astrocyte end-feet were smaller than the gaps between nanofibers. The estimated distance between iPS-ECs and cell bodies of the closest layer of iPS-Astros (attached to the nanofiber layer) was ~2 μm, in good agreement with thickness of the nanofiber layer.

Figure 4.

(a-c) Confocal images of the paper/nanofiber-based BBB model showing the morphology of (a) iPS-ECs, (b) iPS-Astros and (c) merged cells. (d) 3D rendered cross-section image of the BBB model showing the location and distribution of cell components. Scale bars: 100 μm.

To demonstrate the advantage of employing the paper/nanofiber-based cell culture platform to build the BBB model, iPS-ECs and iPS-Astros were co-cultured on a porous PET membrane transwell to generate a traditional in vitro BBB model to compare with the paper/nanofiber-based BBB model (Figure S9). In the traditional transwell model, both iPS-ECs and iPS-Astros form 2D layers, separated by the thick (~25 micron) polymer membrane. Comparatively, in the paper/nanofiber platform developed here, iPS-ECs and iPS-Astros are in much closer physical proximity, with the 3D end-feet features of iPS-Astros being fully developed. These results indicate that the paper/nanofiber-based BBB model reproduces spatial arrangements more similar to those observed in vivo, and based on results described in detail below, results in conditions favorable for enhanced interactions between endothelial and astrocyte layers.

2.5. Barrier function of the BBB model

The barrier properties of the paper/nanofiber-based BBB model were first evaluated by detection of tight junction proteins with immunolabeling and subsequent confocal imaging. As shown in Figure 5a and Figure S10, iPS-ECs displayed positive signals for TJ protein 1 (ZO-1), which is a major tight junction protein found in brain endothelial cells.^[62] Transendothelial electrical resistance (TEER) measurements were then performed to characterize the model. TEER values were recorded for iPS-ECs on the platform alone and for iPS-ECs + iPS-Astros (complete BBB model) co-cultured on the platform (Figure 5b). The empty platform (nanofiber/paper and Matrigel) had a TEER value of 15 ± 4 Ω·cm² (n = 3), revealing that the Matrigel-coated nanofiber layer produced a minimal transport barrier. For the case of iPS-ECs cultured alone on the platform, a TEER value of 165 ± 25 Ω·cm² (n = 3) was measured, in good agreement with other reports of iPS-ECs monolayers.^{[59, 63, 64]} Addition of iPS-Astros in co-culture increased TEER values to 230 ± 15 Ω·cm² in the BBB model (n = 7, p-value = 0.01), demonstrating positive regulation of resistance by astrocytes.

Figure 5.

Evaluation of the barrier function of the paper/nanofiber-based BBB model. (a) Confocal images of the BBB model immunolabeled with ZO-1 showing the distribution of tight junction proteins. DAPI was used to label the nuclei. Scale bars: 100 μm. (b) TEER value of the empty substrate without cells, the culture platform with iPS-ECs alone and with co-culture of iPS-ECs and iPS-Astros. (c) Permeability coefficient of sodium fluorescein through the empty substrate, the empty transwell membrane and the BBB model. (d) Addition of 0.5 M mannitol in the apical side induced TEER loss of the BBB model.

Permeability assays of sodium fluorescein were also carried out on the empty paper culture platform, empty PET transwell membrane, and the paper/nanofiber-based BBB model (Figure 5c). The permeability coefficient of the empty platform was 38 ± 3 × 10⁻⁶ cm/s (n = 3), which is significantly higher (p-value = 0.001) than the transwell membrane with 13 ± 1 × 10⁻⁶ cm/s (n = 3), supporting the assertation that transport of ions and molecules through the platform developed here is enhanced relative to typical membranes used for cell culture. After the BBB model was formed on the platform, permeability decreased to 2 ± 2 × 10⁻⁶ cm/s (n = 4, p-value << 0.001 compared to empty platform), which is similar to other reports.^[43]

To further verify the BBB phenotypes, the BBB model was treated with mannitol and the TEER value response monitored. Mannitol is a hyperosmolar agent used in the treatment of brain trauma and is widely known to diminish the barrier function of the BBB.^[65–67] Mannitol was introduced at the apical side of the BBB model and a clear decrease of the TEER value was observed (Figure 5d). After 90 min of mannitol treatment, the TEER value dropped by ~40% of the original value, showing the paper/nanofiber-based BBB model mimics observations for in vivo BBB to hyperosmotic conditions.

The paper/nanofiber-based BBB model used here displayed significantly lower TEER value than the in vivo BBB (> 5000 Ω·cm²).^[68] This observation of lower barrier function is also commonly found in other in vitro BBB models.^{[63, 64]} Several approaches can be implemented to improve the barrier properties, including the use of primary endothelial cell culture^[69] and retinoic acid treatment of iPS-ECs^[60] bringing the TEER values of in vitro BBB models close to in vivo condition. In addition, the paper/nanofiber-based BBB model possesses similar impaired barrier properties to in vivo BBB under hyperosmotic condition, demonstrating the ability of paper/nanofiber-based BBB model in representing in vivo BBB under certain conditions.

2.6. RNA-sequencing analysis

Generation of brain microvascular endothelial cells (BMECs) from hiPSCs has been explored extensively in recent years.^[70] In particular, generating a BMEC model proves somewhat fickle and is dependent on a number of factors including seeding density,^{[71, 72]} timing of differentiation,^{[59, 73]} influence of additional cells in co-culture,^{[63, 74]} surface dimensionality^{[75, 76]} and fluid shear.^{[70, 74]} Most commonly, quantitative PCR analysis has been used to identify select proteins of interest to determine BMEC differentiation.^{[19, 60, 71, 77]} Recently, RNA-sequencing and single-cell RNA-sequencing have been used to more thoroughly assess transcriptional profiles.^[78–80] Here, to analyze the genetic consequences of the 3D BBB model in a comprehensive fashion, total RNA was extracted from i) the 3D BBB model on the cell culture platform, ii) iPS-ECs alone on a Matrigel surface, and iii) iPS-Astros alone on a Matrigel surface. High-throughput RNA-sequencing analysis (RNA-seq) was then performed to investigate the gene profiles from these three conditions, where the expression level of BBB-related genes in the 3D BBB model was compared to iPS-ECs cultured alone to demonstrate the effect of cell culture platform and astrocytes on the functionality of BMECs.

Figure S11a shows the principal component analysis (PCA) generated from DESeq2 package of the gene profiles collected from all samples, which statistically indicates the similarity between each individual sample analyzed, with PC1 and PC2 accounting for 83% and 16% of variance, respectively. Clustering for each set of triplicate measurements of samples (BBB model, iPS-ECs, and iPS-Astros), indicates minimal variation among replicants within a sample, as compared to across the samples. Gene profiles between different conditions exhibited distinguishable differences, which could also be visualized in a Venn diagram (Figure S11b) summarizing the number of differentially expressed genes in each comparison (BBB model vs iPS-ECs, BBB model vs iPS-Astros, iPS-ECs vs iPS-Astros) at < 5% false discovery rate (FDR) and |log2 fold change| > 2. A volcano plot (Figure S11c) was used to identify the detailed differential expression of genes in the BBB model vs iPS-ECs comparison. In total, 1886 genes were identified as up-regulated and 755 genes were identified as down-regulated in the BBB model, as compared to iPS-ECs alone (|log₂ fold change| > 2, FDR < 5%). For comparison, variance in gene profiles between the BBB model was compared to both iPS-ECs and iPS-Astros alone. Differentially expressed genes were examined between the BBB model and iPS-ECs for characterization of functional barrier and BBB phenotypes, and differentiation of endothelial vs. epithelial compositions present in iPS-ECs. BBB-related genes in the composite were found to be uniquely regulated in the BBB model, supporting conclusions discussed in detail below.

Figure 6a shows a clustering heatmap of 48 genes with relevance to the BBB as expressed in the BBB model, iPS-ECs and iPS-Astros, including endothelial markers, astrocyte markers, tight junction markers, and membrane transporters, as well as ECM and ETS transcription factors. All astrocyte markers (GFAP, S100B, ACTA2) had a strong contrast in the expression level between iPS-Astros and iPS-ECs and were put in the highest hierarchy of the heatmap, which further proved the high quality of the iPS-Astros. Although endothelial marker PECAM-1 protein was found abundant in the BBB model by immunolabeling (Figure 4), gene expression of PECAM1 was statistically lower (p-value = 0.002) in the BBB compared to the iPS-ECs alone. Genes of ECM proteins were also found at a lower expression level, possibly due to the enrichment of ECM at the nanofiber layers, which lowers the need of cells to express ECM for proper cell-matrix interactions. For other endothelial markers and most of the genes highly expressed in iPS-ECs, the BBB model exhibited a similar or even higher level (explained more below). This implies that the use of the cell culture platform did not impair the properties of iPS-ECs related to endothelial functions.

Figure 6.

(a) Clustering heatmap of 48 genes related to different BBB characteristics and functions expressed in the BBB model, iPS-ECs and iPS-Astros. (b) BBB-related genes that showed significant regulation in the BBB model compared with both iPS-ECs and iPS-Astros at 5% FDR (*p < 0.05, **p < 0.01, ***p < 0.001).

Figure 6b summarizes the expression levels of the genes from the heatmap that were regulated uniquely in the BBB model, meaning that these genes were up-regulated or down-regulated compared to both iPS-ECs and iPS-Astros alone at 5% FDR. Endothelial markers VWF, KDR, ENG, which are responsible for blood hemostasis,^[81] mediation of vascular endothelial growth factor,^[82] and angiogenesis,^[83] respectively, were identified as up-regulated in the BBB model. Up-regulation of these genes supports the notion that introduction of iPS-Astros could help iPS-ECs develop endothelial phenotypes related to blood vessel stability and permeability. Two ATP-binding cassette transporters important for BBB functions, ABCG2 and ABCB1, were also differentially regulated in the BBB model. While ABCG2 representing breast cancer resistance protein (BCRP) displayed a 7-fold increase, ABCB1 (P-Glycoprotein) was decreased by 80% in the BBB model, which is opposed to previous reports of astrocytes increasing functional expression of P-glycoprotein in an in vitro BBB model.^[84] As P-Glycoprotein is a gatekeeper of the BBB^[85] and BCRP mediates multidrug resistance,^[86] regulation of these genes may indicate altered transport selectivity of iPS-ECs in the presence of iPS-Astros and the cell culture platform. The VE-Cadherin gene, CDH5, was found to have a 3-fold increase in the BBB model compared to iPS-ECs alone. VE-Cadherin is an important component of endothelial cell adherens junctions and has been implicated in playing a fundamental role in controlling the transport across the endothelial barrier and in regulating angiogenesis.^{[87, 88]} CLDN2, a member of the claudin family, was found to be down-regulated in the BBB model. As CLDN2 expresses paracellular water channels formed at the tight junction complexes,^[89] depletion of CLDN2 increases barrier properties of paracellular spaces leading to increase of TEER.^[90] Figure S12 also displays the expression level of other adherens junction and tight junction genes, including CDH1, CDH3, CLDN4, CLDN5, TJP1, TJP2, where the BBB model exhibited similar or slightly higher expression levels than iPS-ECs alone. These gene expression profiles support the assertation that increase of TEER observed in the BBB model compared to iPS-ECs could be attributed to the regulation of junctional genes, including CDH5 and CLDN2, induced by the introduction of iPS-Astros and closer endothelial-astrocyte interactions provided by the cell culture platform.

Recently, Lu, et al. suggested that the method used here, and widely employed in a wide number of BBB studies, to differentiate hiPSCs into iPS-ECs^[59] results in differentiated cells with epithelial-like characteristics.^[91] Comparison of RNA-seq data collected here with that report showed iPS-ECs used in this study indeed expressed epithelial genes at a high level, including CLDN4, EPCAM, ERBB3 and ESRP1 (data not shown). As pointed out by Lu, et al., iPS-ECs lack the ability to form proper vascular structure, and recapitulation of certain vascular functions and endothelial phenotypes could be obtained by overexpression of three ETS transcription factors: ETV2, FLI1, and ERG.^[92] Therefore, expression level of ETV2, FLI1, and ERG were compared here, between the BBB model and iPS-ECs alone. While ETV2 and ERG did not show significant change in the BBB model (Figure 6a), FLI1 exhibited a 4.5-fold increase in the BBB model compared with iPS-ECs (Figure 6b). A considerable increase of FLI1 in the BBB may demonstrate that the introduction of iPS-Astros could rescue iPS-ECs with improvements of endothelial cell type related to vasculogenesis,^[93] although further work is necessary to validate this assertation. Furthermore, as discussed above, up-regulation of endothelial markers, VWF, KDR, and ENG support endothelial properties of the transformed cells when in the BBB model. Gene profiling described here is consistent with a mixed endothelial-epithelial transcriptional profile, indicative of brain microvascular endothelial-like cells,^[70] with the caveat that the co-cultured BBB model developed here shows improvement of key endothelial makers relative to hiPSCs cultured alone.

3. Conclusion

In this study, we have demonstrated a novel 3D BBB model based on a hybrid paper/nanofiber-based cell culture platform. The platform consists of a high-porosity lens paper sheet with a Matrigel-coated nanofiber layer on the apical side. A co-culture of hiPSC-derived iPS-ECs and iPS-Astros on the platform was used to develop a more representative human BBB model for biomedical applications. With the paper scaffold, iPS-Astros successfully formed 3D morphologies with end-feet features. iPS-ECs and iPS-Astros were also in closer proximity in the paper platform compared to the traditional 2D transwell membrane, which could facilitate the endothelial-astrocyte interaction. The paper/nanofiber-based BBB model possessed high-quality barrier functionality, as evaluated by TEER measurements and permeability assays, and exhibited disruption in response to mannitol treatment akin to a bona fide BBB. The addition of iPS-Astros increased the barrier properties of iPS-ECs indicating positive regulation of astrocytes to the BBB model. Further RNA-seq gene analysis revealed that the introduction of iPS-Astros regulated the gene expression of the tight junction protein family and VE-cadherin in iPS-ECs, explaining the increased TEER observed. While most of the BBB-related transporters had been found abundantly expressed in the BBB model, several transporters exhibited significant change in expression level indicating altered transport selectivity in the endothelial barrier. Although iPS-ECs used in this study have recently been described as presenting a more epithelial-like phenotypes and possibly lacking the potential to form blood vessels, the introduction of iPS-Astros greatly increased one of the ETS transcriptor, FLI1, known to rescue the iPS-ECs back to a vascular endothelium cell type. In addition, a majority of endothelial markers were found to be up-regulated in the BBB model, as compared to iPS-ECs alone. The paper/nanofiber-based BBB model has clearly provided a new approach to generate a 3D BBB structure with low cost and ease of large-scale manufacturing, which is beneficial for high-throughput drug screening for efficient drug delivery. In addition, the paper/nanofiber-based cell culture platform has the potential to be used to study the interaction between any two cell lines (e.g., fibroblasts and epithelial cells, neurons and astrocytes), extending its applications in building other 3D tissue models.

4. Experimental Section/Methods

Fabrication of the cell culture platform:

A polystyrene (PS) nanofiber layer was electrospun directly onto commercially available high-porosity paper by first dissolving 12% (w/v) polystyrene pellets (280 kDa M.W., Sigma-Aldrich, MO, USA) and 1% Tetrabutylammonium bromide (Sigma-Aldrich) in dimethylformamide (DMF) on a shaker for at least 2 hrs. The solution was then added to a glass vial with a septum and pressurized at 3–5 psi with helium to induce flow through a 15 cm, 150 μm I.D. fused silica capillary. A platinum wire was used to make electrical contact and a voltage of 12–14 kV was applied to the solution. An 11 cm x 20 cm sheet of Whatman lens cleaning tissue, Grade 105 (Sigma-Aldrich) was wrapped around a grounded drum collector (11 cm long by 6 cm in diameter) that rotated at ~300 rpm while also sliding horizontally over a 12 cm span at ~6 cm/ second. This target was placed 20 cm from the end of the fused silica capillary. Electrospinning was performed in a custom plexiglass enclosure at an ambient temperature of 22 °C and humidity of under 20%. Pictures of the set up are shown in Figure S2. Fibers were collected at discrete points from 10 to 90 minutes depending on the desired scaffold thickness.

The hybrid nanofiber/paper scaffold was taped to a customized transwell (previously reported^[90]) and plasma cleaned for 10 s at medium RF level to make the surface hydrophilic. Matrigel without growth factor (Corning, Corning, NY, USA) was added to uniformly cover the nanofiber layer at 200 μg/cm² and then air dried. The cell culture platform was sterilized under UV light for at least 2 h before use. For SEM characterization, the platform was sputter-coated with a 1 nm Au/Pd layer and then imaged with an FEI Quanta 600 scanning electron microscope. Surface topography of the platform was also measured by a customized SICM instrument where the platform was immersed in phosphate buffered saline (PBS).

iPS cell culture and differentiation:

iPS-ECs were derived from the modified version of a previously published method.^[59] Briefly, hiPSCs (Coriell Institute, NJ, USA) were sub-cultured onto a Matrigel-coated flask at 50,000 cells/cm² and maintained in mTeSR1 medium (STEMCELL Technologies, Vancouver, Canada) for 3 days prior to differentiation. The medium was then switched to unconditioned medium (UM) consisting of 1:1 Dulbecco’s Modified Eagle’s Medium/Ham’s F-12 Medium (DMEM/F12, Life Technologies, CA, USA) with 20% Knockout Replacement Serum (Life Technologies), 1x MEM nonessential amino acids (Sigma), 1 mM L-glutamate (Life Technologies) and 0.1 mM β-mercaptoethanol (Sigma) to initiate differentiation. After 8 days with the medium refreshed every other day, UM was replaced with endothelial cell media (EM) consisting of human endothelial serum-free medium (Life Technologies), 1% human platelet-poor plasma-derived serum (Sigma) and 20 ng/mL human basic fibroblast growth factor (bFGF, Sigma). After 3 days of EM treatment, the resultant iPS-ECs were dissociated with Accutase (Life Technologies) and seeded onto the desired substrate (cell culture platform or transwell membrane). iPS-Astros were obtained from the maturation of iPSC-derived astrocyte precursor cells (Applied StemCell, CA, USA). Briefly, precursors were seeded onto Matrigel and maintained in astrocyte maturation medium (Applied StemCell) for 7 days with medium refreshed every other day. Cells were then passaged onto a new Matrigel surface and maintained under the same conditions for another 7 days.

Building the BBB model:

Before cell seeding, EM was added to the cell culture platform to rehydrate the Matrigel-coated nanofiber layer. The platform was then coated with a mixture of collagen IV (400 μg/mL, Sigma) and fibronectin (100 μg/mL, Sigma) and incubated with the coating for 4 h at 37°C. After incubation, Matrigel coating was added to the basal side of the platform (paper). iPS-Astros were disassociated with Accutase, resuspended with astrocyte medium composed of DMEM with 10% fetal bovine serum (FBS), 1x N2 supplement (Life Technologies), 1% GlutaMAX (Life Technologies), 20 ng/mL bFGF, and then seeded at 10⁶ cells/mL in the paper. After 4–6 h to allow attachment of iPS-Astros to the paper, iPS-ECs were seeded onto the nanofiber layer at 2 × 10⁶ cells/mL. iPS-ECs and iPS-Astros were maintained in EM and astrocyte medium respectively with media refreshed every other day for 7 days to generate the BBB model used for subsequent analysis.

Cell viability assay:

The viability of iPS-ECs and iPS-Astros cultured on the platform was evaluated by calcein AM imaging. Briefly, the cells were rinsed with PBS and then incubated with 5 μM calcein AM (AAT Bioquest, CA, USA) for 30 min, followed by PBS washing. Fluorescence images of the cells were obtained with a Nikon E800 optical microscope.

Immunofluorescence imaging:

The sample (the BBB model, iPS-ECs or iPS-Astros) was first rinsed with PBS, followed by fixation and permeabilization with methanol at −20°C for 20 min and twice washed with PBS for 5 min. Primary antibodies (1:100, volume ratio) conjugated to fluorophore against PECAM-1 (conjugated to Alexa546, Santa Cruz Biotechnology, CA, USA) and GFAP (conjugated to Alexa488, Life Technologies) were used to identify iPS-ECs and iPS-Astros, respectively. To characterize the presence of tight junctions, primary antibodies were used against ZO-1 (conjugated to Alexa488, Santa Cruz Biotechnology). The sample was incubated with primary antibodies for 1–2 h at RT, and then with 4′,6-diamidino-2-phenylindole (DAPI, Life Technologies) for 2 min to label the nuclei. After PBS washing, the sample was imaged by a Leica SP8 confocal laser-scanning microscope. Post-image processing and 3D rendering were performed by ImageJ and Imaris (Bitplane, Switzerland), respectively.

TEER measurement:

The electrical resistance of the empty cell culture platform, iPS-ECs on platform alone and the BBB model were measured by STX2 chopstick electrodes and a EVOM2 Voltohmmeter (World Precision Instruments, FL, USA) and normalized to the TEER value with exposed surface area as follows:

$$\mathrm{TEER}(\mathrm{\Omega }·\mathrm{c}{\mathrm{m}}^2)=\mathrm{resistance}\left(\mathrm{\Omega }\right)\mathrm{x}\mathrm{exposed}\mathrm{surface}\mathrm{area}(\mathrm{c}{\mathrm{m}}^2)$$

Sodium fluorescein permeability assay:

Sodium fluorescein permeability studies were performed for the empty cell culture platform, 0.4 μm pore size transwell membranes (Corning), and the BBB model. After PBS washing, 3 mL of 1 mM sodium fluorescein in PBS solution was added to the apical side of the sample while the basal side was refilled with 4 mL PBS (without sodium fluorescein). After 30 min, the solution on the basal side was collected and the concentration of sodium fluorescein was determined by a Varian Cary Eclipse fluorescence spectrometer. The permeability coefficient, P, can be calculated as follows:

$$\mathrm{P}=\frac{\Delta Q}{C\times A}=\frac{C_{Basal}\times V_{Basal}}{t\times C\times A}$$

where Q is the volumetric flow rate of the solution, C_Basal is the concentration of sodium fluorescein in the basal side, V_Basal is the volume of solution in the basal side, t is the measurement time, C is the original concentration of sodium fluorescein in the apical side, and A is the exposed surface area.

Mannitol effect study:

To study the effect of mannitol on the barrier function of the BBB model, EM in the apical side of the BBB model was replaced with a fresh EM supplement containing 0.5 M mannitol (Osmolality = 800 mOsmol per kg H₂O). The change of TEER was monitored for the next 90 min with impedance spectroscopy performed on a CHI660C electrochemical workstation (CH Instruments, TX, USA). A control experiment was also conducted under the same conditions except that no mannitol was present in the system.

RNA-seq experiments:

Total RNA of iPS-ECs, iPS-Astros and the BBB model were isolated with TRIzol reagent (Life Technologies). Before the construction of the library for RNA-seq, all RNA samples were evaluated on an Agilent TapeStation 3000 and only samples with RIN score larger than 9 were used. An aliquot of 300 ng of total RNA was used for each sample and ddH₂O was added to bring the final volume to 25 ul. Sequencing libraries were prepared with Illumina strand-specific mRNA library preparation Kit following the standard company protocol with half reaction for each step. The final library concentration and distribution was checked on an Agilent D1000 screen tape. Based on the TapeStation molarity value, libraries were pooled and sequenced on a NextSeq 500 sequencing run generating PE 42 bp long reads. Reads were adapter trimmed and quality filtered using Trimmomatic ver. 0.38^[94] setting the cutoff threshold for average base quality score at 20 over a window of 3 bases, excluding the reads shorter than 20 bases post-trimming (parameters: LEADING:20 TRAILING:20 SLIDINGWINDOW:3:20 MINLEN:20). Cleaned reads were aligned to the human genome reference sequence GRCh38.p12 using STAR version 2.7.3a.^[95] Read pairs mapping concordantly and uniquely to the exon regions of the annotated genes were counted using featureCounts tool ver. 2.0.0 of subread package.^[96] Read alignments to antisense strand, or to multiple regions on the genome or those overlapping with multiple genes were ignored (parameters: -s 2 -p -B -C). Differential expression analysis was performed using DESeq2 ver. 1.24.0^[97] and the p-values were corrected for multiple-testing using the Benjamini–Hochberg method.