Generation of Brain Microvascular Endothelial-Like Cells from Human Induced Pluripotent Stem Cells by Co-Culture with C6 Glioma Cells

The blood brain barrier (BBB) is formed by brain microvascular endothelial cells (BMECs) and tightly regulates the transport of molecules from blood to neural tissues. In vitro BBB models from human pluripotent stem cell (PSCs)-derived BMECs would be useful not only for the research on the BBB development and function but also for drug-screening for neurological diseases. However, little is known about the differentiation of human PSCs to BMECs. In the present study, human induced PSCs (iPSCs) were differentiated into endothelial cells (ECs), and further maturated to BMECs. Interestingly, C6 rat glioma cell-conditioned medium (C6CM), in addition to C6 co-culture, induced the differentiation of human iPSC-derived ECs (iPS-ECs) to BMEC-like cells, increase in the trans-endothelial electrical resistance, decreased in the dextran transport and up-regulation of gene expression of tight junction molecules in human iPS-ECs. Moreover, Wnt inhibitors attenuated the effects of C6CM. In summary, we have established a simple protocol of the generation of BMEC-like cells from human iPSCs, and have demonstrated that differentiation of iPS-ECs to BMEC-like cells is induced by C6CM-derived signals, including canonical Wnt signals.


Introduction
The blood brain barrier (BBB), which is formed by specialized brain microvascular endothelial cells (BMECs) working together with astrocytes and pericytes, plays important roles in brain homeostasis and neuronal functions by regulating the transit of substances from peripheral circulation to brain. Unlike endothelial cells (ECs) that reside in the other tissues or organs, BMECs highly express genes associated with tight junction molecules and efflux/influx transporters, and thereby could regulate the entrance of various types of compounds such as small molecules and drugs, into the brain [1]. To analyze the function of BBB and to examine the permeability of compounds through the BBB, in vitro BBB models have been developed using cultured BMECs that were mostly derived from non-human animals [1,2]. However, since the expression pattern and the expression levels of transporters are different between non-human animals and human [3,4], establishment of in vitro BBB models using human BMECs would be preferable. Nonetheless, the use of human primary BMECs also has some drawbacks. One is their limited range of sources and differences in function from batch to batch. In addition, although immortalized human BMECs have been established by the transduction of tumor genes including SV40 large T-antigen, these cells show lower barrier functions compared to primary BMECs in general [5,6].
Human pluripotent stem cells (PSCs), such as embryonic stem cells [7,8] and induced PSCs (iPSCs) [9,10], can differentiate into various types of cells in the body in an unlimited quantity. Human PSC-derived ECs are therefore expected to be used as sources for human BMECs. Many researchers have reported the differentiation of human PSCs into ECs using various strategies [11][12][13][14]. However, few studies investigated the generation of tissue-specific ECs, including BMECs, from human PSCs.
Since BBB is formed simultaneously with brain development, it was reasonably assumed that BMECs would be maturated by stimulation of the factors produced by other types of cells including neural tissue-related cells. In this study, we initially attempted to establish a method for the differentiation of human iPSCs into ECs under serum-and feeder-free conditions, and examined whether human iPSC-derived ECs (iPS-ECs) could be further maturated to BMEClike cells by co-culture with several kinds of cultured cell lines. In addition, the effects of cell line-derived conditioned medium on the differentiation of iPS-ECs into BMEC-like cells were also examined to establish a simple protocol for the generation of BMEC-like cells from human iPSCs.

Differentiation of ECs from human iPSCs under a serum-free condition
To generate brain-specific ECs from human iPSCs, human iPSCs were initially differentiated into ECs under serum-free conditions as illustrated in Fig 1A. Flow cytometric analysis revealed that the degree of CD34 + CD144 + endothelial progenitor cells in EB cells was increased to a peak on day 9, and decreased over the next 3 days (Fig 1B). We also observed the enhanced expression of EC-related genes in EBs on day 9 compared to those in EBs on day 6 ( Fig 1C). The expression of PSC-marker genes, Nanog and Oct-3/4, was markedly decreased following EB cultures (Fig 1C). Thus, under our culture conditions, a large number of endothelial lineage cells were observed in 9-day-cultured EBs. To obtain pure ECs, we isolated CD34 + CD144 + cells using anti-human CD34 antibodies, because all CD144 + cells highly expressed CD34 ( Fig  1B). After cell sorting, EC-like cells were observed from CD34 + cell fractions in fibronectincoated plate culture (Fig 1D), and these cells expressed CD31 and vWF ( Fig 1E). Moreover, in these cells, a vascular-like structure was formed in Matrigel ( Fig 1F) and uptake of the acetylated LDL was observed (Fig 1G). By contrast, CD34cell-derived cells did not express CD31 and vWF, and did not show tube formation potential (S1 Fig). Therefore, these results clearly demonstrated that functional ECs were generated from human iPSCs under serum-and feederfree conditions. Human iPSC-derived EBs (day 6, day 9 and day 12) were stained with anti-CD34 and anti-CD144 antibodies, and were then subjected to flow cytometric analysis. Representative results from one of the three independent experiments are shown. (C) Total RNA was extracted from undifferentiated human iPSCs (day 0) and human iPSC-derived EBs (day 6 and day 9). Then, qRT-PCR analysis was performed. Results shown are the mean of three independent experiments with the indicated standard deviations (S.D.). * p < 0.05, ** p < 0.01. (D-G) Sorted CD34 + cells were cultured with FGF2, ECGS, and heparin on fibronectin-coated plates. They showed an endothelial-like morphology under these culture conditions (D). These cells were stained positive for CD31 and vWF (E). They were also capable of forming tube-like structures on Matrigel (F) and demonstrated acetylated-LDL uptake (G). The scale bar indicates 300 μm (D) or 100 μm (E, F, G).

Induction of BMEC-like cells from human iPS-ECs by co-culture with C6 cells
Recent studies have indicated the pivotal roles of tissue microenvironments in the acquisition of tissue-specific properties of ECs [15]. In addition, iPS-ECs were considered to be "naïve" or "uncommitted" ECs, probably because they were generated and expanded in the absence of tissue-specific factors. Thus, it is expected that co-culture of "naïve" iPS-ECs with neural tissuerelated cell lines would lead to the generation of BMEC-like cells with tight junction-forming potential. We screened cultured cell lines that have activities potentiating the barrier function of ECs using b.End3 cells (a mouse BMEC line), and found that rat C6 glioma cells have the strongest capacity to promote the tight junction formation of b.End3 cells (S2 Fig). Thus, BMEC-like cells were generated from iPS-ECs by co-culture with C6 cells. HUVECs were also co-cultured with C6 cells as a control.
The TEER value of iPS-EC monolayers (iPSEC-mono) was 22-28 O cm 2 and was unchanged after the culture on Transwell inserts, whereas that of iPS-ECs co-cultured with C6 cells (iPSEC-C6) was gradually increased and peaked on day 5 (approximately 55 O cm 2 ) ( Fig  2A). We also observed a significant elevation of gene expression of tight junction-related genes, Claudin-5, Occludin, and Zonula occludens-1 (ZO-1), in iPSEC-C6 ( Fig 2B). Furthermore, coculture with C6 cells led to a decrease in dextran (MW 3,000) transport in iPS-ECs ( Fig 2D). These data thus indicate that iPSEC-C6 could form tight junctions and thereby could have barrier functions. By contrast, the barrier functions of HUVECs were not altered even in the presence of C6 cells (Fig 2A-2D).
To provide nutrients to the brain and to protect the brain from exposure to unnecessary metabolites and drugs, BMECs express various types of influx and efflux transporters. Thus, the expression levels of transporter genes in iPSEC-mono and iPSEC-C6 were investigated by RT-PCR. The results showed that, compared with iPSEC-mono, iPSEC-C6 showed an increased expression of Mdr1, Bcrp, Mrp-4 and Glut1 (Fig 2E). We next examined the function of a transporter p-gp in iPSEC-C6 using rhodamine123, a substrate for p-gp, and cyclosporine A, an inhibitor of p-gp. As shown in Fig 2F, the transport of rhodamine123 from the upper chambers to the lower chambers was slightly but significantly increased by treatment with cyclosporine A, but not by treatment with MK571, an inhibitor of the MRP family, indicating that iPSECs-C6 could express functional p-gp. Together, these results suggest that iPS-ECs would be maturated to BMEC-like cells, which form tight junctions and express various types of transporters, by co-culture with C6 cells.

Maturation of iPS-ECs to BMEC-like cells using C6 cell-conditioned medium (C6CM)
Considering that C6 cells did not directly contact iPS-ECs under our culture conditions, the results in Fig 2 suggest that C6 cell-derived soluble factors promote the maturation of iPS-ECs to BMEC-like cells. We thus investigated the effect of C6CM on the maturation of iPS-ECs. Interestingly, TEER measurement, gene expression analysis and dextran transport assay showed the increased barrier properties of iPS-ECs in the presence of C6CM (iPSEC-C6CM), as in the case of co-culture with C6 cells (Fig 3A-3C). Furthermore, addition of C6CM to the cultures resulted in the enhanced expression of transporter genes in iPS-ECs ( Fig 3D). We also confirmed the functionality of p-gp expressed in iPSECs-C6CM ( Fig 3E). Therefore, C6 cell-derived soluble factors would be sufficient to induce maturation of iPS-ECs to BMEC-like cells.

Fig 2. Induction of BMEC-like properties in hiPS-ECs by co-culture with C6 cells. (A) hiPS-ECs and
HUVECs were cultured on fibronectin-coated inserts. When confluent, these inserts were transferred onto non-cultured 24-well plates (iPSEC-mono or HUVEC-mono) or C6 cell-cultured 24-well plates (iPSEC-C6 or HUVEC-C6). Then, TEER values for hiPS-ECs and HUVECs were measured at indicated days. (B, C) Expressions of tight junction-related genes (Claudin-5, Occludin and ZO-1) were examined by qRT-PCR analysis before (iPSECs, HUVECs) and after (iPSEC-mono, iPSEC-C6, HUVEC-mono, HUVEC-C6) 5 days of culture. (D) The permeability coefficient for FD was investigated in 5-day co-cultured hiPS-ECs and HUVECs. (E) Expressions of transporter genes (P-gp, Bcrp, Mrp-4 and Glut1) were examined in 5-day mono-or co-cultured hiPS-ECs by RT-PCR analysis. (F) The 5-day co-cultured hiPS-ECs were treated with CSA (an inhibitor of P-gp) or MK571 (an inhibitor of MRP-4), and then the permeability coefficient for Rhodamin 123 (a specific substance of P-gp) was measured. All results shown are the mean of three independent experiments with the indicated standard deviations (S.D.). * p < 0.05. Involvement of Wnt signaling in the maturation of iPS-ECs to BMEC-like cells Canonical Wnt signaling via β-catenin has been reported to contribute to angiogenesis in the brain [16][17][18]. Therefore, we expected that C6CM would stimulate the canonical Wnt pathway in iPS-ECs, and the activated Wnt signaling would lead to the maturation of iPS-ECs to BMEC-like cells. Indeed, the expression of Axin2, a target gene of ß-catenin, was increased in iPSEC-C6 (Fig 4A), suggesting activation of the canonical Wnt pathway. We studied the involvement of canonical Wnt signaling in the maturation of iPS-ECs using inhibitors (Dkk1 and XAV931), and found that the inhibition of canonical Wnt signaling by Dkk1 and XAV931 led to a slight decrease in the barrier functions of iPSECs-C6CM, although Dkk1-or XAV931-treated iPSEC-C6CM still displayed barrier functions superior to those of iPSECmono (Fig 4B and 4C). Thus, these results suggest that C6 cell-induced maturation of iPS-ECs to BMEC-like cells would be partially mediated by canonical Wnt signaling.

Discussion
The aim of this study is to generate human BMECs from PSCs, because an establishment of in vitro BBB models using human BMECs would be required to analyze the BBB functions and to perform drug-screening for neurodegenerative diseases. In the present study, we initially succeeded in the generation of iPS-ECs under serum-and feeder-free conditions (Fig 1). Furthermore, we showed that the TEER values in iPSEC-C6 were higher than those in iPSEC-mono, HUVEC-mono, and HUVEC-C6 cells, and that paracellular transport was significantly reduced in iPSEC-C6, indicating the increased barrier functions (Fig 2). In addition, we found an elevation of the expression level of genes involved in influx and efflux transporters in iPS-ECs by co-culture with C6 cells (Fig 2). The TEER values (~55 O cm 2 ) in iPSEC-C6 were largely identical to those in hCMEC/D3 cells (30~40 O cm 2 ) [5], which are immortalized human BMECs that have been widely utilized as cell sources for human in vitro BBB models [19,20]. Our results, therefore, indicate that iPS-ECs would be maturated into BMEC-like cells by coculture with C6 cells.
C6 cells have the properties of astrocytes and maintain the functionality of primary BMECs in vitro [21,22]. It has also been shown to increase in the barrier functions of primary BMECs in vitro by treatment of C6CM [23]. In addition to these functions, we found, for the first time, that C6 cells could facilitate the maturation of "naïve" iPS-ECs into BMEC-like cells. On the other hand, C6 cells had almost no impact on the barrier functions and the expression levels of transporter genes in HUVECs. Because HUVECs were derived from "mature" or "committed" ECs in the umbilical cord, they might not be converted into BMEC-like cells even in the presence of C6 cell-derived factors. These findings suggest the possibility that both "naïve" ECs and signals derived from neural tissue-related cells would be necessary to induce the BMEC-like cells. Considering the difficulty in obtaining a large amount of human primary BMECs and the difficulty of conversion of tissue-or organ-derived mature ECs into BMECs, iPS-ECs would be an ideal cell source for human BMEC generation. In the future, therefore, it will be important to investigate whether ECs derived from other kinds of iPSCs could differentiate into BMEClike cells by co-culture with C6 cells.
Another important finding in this study was that C6CM could also promote the maturation of iPS-ECs into BMEC-like cells. These data suggest that C6 cell-derived soluble factors would play an important role in the maturation of iPS-ECs. Consistent with the previous findings, we confirmed that canonical Wnt signaling contributed to the maturation of iPS-ECs by C6CM treatment. However, not only the canonical Wnt pathway but also other unknown factors are likely to be involved in the C6CM-elicited maturation of iPS-ECs, since treatment of canonical Wnt inhibitors did not completely block the maturation of iPS-ECs. The identification of C6 cell-derived maturation factors in the C6CM would be important to elucidate the maturation mechanisms induced by C6 cells.
Recently, Lippmann et al. showed that BMEC-like cells could be generated from human PSCs by spontaneous differentiation cultures [24,25]. BMEC-like cells generated by their methods had potent barrier functions, but the differentiation efficiencies were assumed to be unstable, due to the dependence on random differentiation. On the other hand, we generated BMEC-like cells from human iPSCs using strategies different from Lippmann's. It would be of interest to compare the characteristics of BMEC-like cells generated from the different protocols.
In summary, we showed that iPS-ECs could be differentiated into BMEC-like cells by coculture with C6 cells or by treatment with C6CM. Maturation levels of BMEC-like cells generated in this study, however, are still insufficient because the TEER values and the expression levels of transporter genes in iPS cell-derived BMEC-like cells are low in comparison with those in primary BMECs [2]. Co-culture of iPS-ECs with human neural tissue-related cells, instead of non-human cell lines, might possibly be effective for generation of fully maturated BMECs from human PSCs. Although further study will be needed to induce the fully maturated BMECs, human iPSC-derived BMECs would be a potential cell source for the generation of in vitro BBB models, and are also expected to be of importance in drug-screening.

RT-PCR
Total RNA was isolated using ISOGEN (Nippon Gene, Tokyo, Japan) or RNAiso Plus (TaKaRa, Shiga, Japan). cDNA was synthesized using a SuperScript VILO cDNA synthesis kit (Life Technologies), and semi-quantitative PCR was then performed using TaKaRa ExTaq HS DNA polymerase (Takara). Semi-quantitative PCR reactions were performed by heating to 94°C for 2 min, and then the samples were subjected to number of cycles of 94°C for 15 sec, 55°C for 30 sec with 72°C for 30 sec and a final extension of 72°C for 1 min. In the case of C6 cell-derived cDNA, 3% dimethyl sulfoxide was added. The product was analyzed by 2% agarose/TBE gel electrophoresis followed by staining with ethidium bromide. Quantitative realtime RT-PCR was performed with Fast SYBR Green Master Mix using an ABI StepOne Plus system (Life Technologies). Relative quantification was performed against a standard curve and the values were normalized against the input determined for the housekeeping gene, glyceraldehyde 3-phosphate dehydrogenase (GAPDH). The sequences of the primers used in this study are listed in Tables 1 and 2.
Tube-forming assay and acetylated low-density lipoprotein (LDL) uptake assay Plates (48-wells) were coated with 100 μl of Matrigel Matrix (BD Bioscience) at 37°C for 1 hr. FACS-sorted cells, which were cultured on fibronectin-coated plates as described above, were Effects of C6 co-culture and C6CM on differentiation Human iPSC-derived CD34 + cells were plated and cultured on fibronectin-coated Transwell inserts (24 well plates) in differentiation medium as described above. When confluent, human iPS-EC monolayers were transferred onto sub-confluent C6 cells in 24-well plates. Then, the cells were further cultured in the EC medium (upper chamber) and differentiation medium (lower chamber). To prepare C6CM, C6 cells were cultured on 10 cm dishes until confluent, and then the medium was replaced with 8 ml of differentiation medium. After 24 hr of incubation, the medium was collected, filtered, and stored at -80°C. When C6CM was used, human iPS-EC monolayers, which were cultured on Transwell inserts until confluent, were transferred onto 24-well plates, and EC medium was loaded onto upper chambers. We added the C6CM (100%) to the lower chambers to mimic the co-culture experiments with C6 cells. In some experiments, 200 ng/ml human Dickkopf1 (Dkk1: Peprotech) or 10 μM XAV939 (Sigma) was added to the cultures. After cells were cultured for the indicated number of days, the characteristics of human iPS-ECs were studied, as described below.

Measurement of trans-endothelial electrical resistance (TEER)
TEER values of human iPS-EC monolayers or HUVEC monolayers on Transwell inserts (24-well plates) were measured using Millicell ERS-2 (Merk-Millipore). To calculate TEER (O (0 ohrms)×cm 2 (the surface area of the insert, 0.3 cm 2 )), the resistance of the fibronectin-coated insert without cells (blank resistance) was always subtracted from the resistance of cell cultures (sample resistance) [2,5] Permeability experiments Human iPS-ECs and HUVECs were cultured on Transwell inserts for 5 days. Before the transport studies, the medium was removed from the upper and lower chambers and the chambers were washed with EBM basal medium phenol red free (Lonza) containing 0.1% FBS (0.1% FBS/EBM). Medium containing 100 μg/ml fluorescein-conjugated dextran (FD, MW 3,000: Life Technologies) was then loaded onto the upper chambers, and 0.1% FBS/EBM was added to the lower chambers. After 15 hr of incubation, the medium in the lower chamber was collected. The concentration of FD in the sample was determined using a fluorescence multiwell plate reader (Ex(λ) 494 nm, Em(λ) 521 nm: Genios, TECAN, Männedorf, Switzerland). The permeability coefficient was calculated as previously reported [26,27].
P-glycoprotein (P-gp) functionality assays P-gp functionality was assessed using rhodamine 123 (Sigma), a substrate for P-gp, and efflux transporter inhibitors as follows. Human iPS-ECs, which were co-cultured with C6 cells or incubated in C6CM, were washed with PBS and pre-incubated at 37°C for 1 hr with or without inhibitors (5 μM cyclosporine A (Wako) or 10 μM MK571 (Sigma)), followed by addition of rhodamine 123 to the upper chamber. After 1 hr, the fluorescence activity of the medium in the lower chamber was quantified on a plate reader [24].

Statistical analysis
Statistical analysis was performed using unpaired two-tailed Student's t-test.