Comparative study of expression and activity of glucose transporters between stem cell-derived brain microvascular endothelial cells and hCMEC/D3 cells

Abstract

Glucose constitutes a major source of energy of mammalian brains. Glucose uptake at the blood-brain barrier (BBB) occurs through a facilitated glucose transport, through glucose transporter 1 (GLUT1), although other isoforms have been described at the BBB. Mutations in GLUT1 are associated with the GLUT1 deficiency syndrome, yet none of the current in vitro models of the human BBB maybe suited for modeling such a disorder. In this study, we investigated the expression of glucose transporters and glucose diffusion across brain microvascular endothelial cells (BMECs) derived from healthy patient-derived induced pluripotent stem cells (iPSCs). We investigated the expression of different glucose transporters at the BBB using immunocytochemistry and flow cytometry and measured glucose uptake and diffusion across BMEC monolayers obtained from two iPSC lines and from hCMEC/D3 cells. BMEC monolayers showed expression of several glucose transporters, in particular GLUT1, GLUT3, and GLUT4. Diffusion of glucose across the monolayers was mediated via a saturable transcellular mechanism and partially inhibited by pharmacological inhibitors. Taken together, our study suggests the presence of several glucose transporters isoforms at the human BBB and demonstrates the feasibility of modeling glucose across the BBB using patient-derived stem cells.

under physiological conditions, the brain in mammalian organisms consumes on average ~20% of the oxygen and 25% of the glucose delivered by the systemic circulation despite its relatively small size (6). Glucose is considered as the main source of energy for the central nervous system under physiological conditions; however, the brain can utilize alternative sources such as ketone bodies as a source of energy (24).

As other solutes, glucose cannot passively diffuse across the blood-brain barrier (BBB) to reach the central nervous system. The BBB is a component of the neurovascular unit formed by specialized brain microvascular endothelial cells (BMECs) surrounded by a basement membrane shared with pericytes, and interacting by close proximity with astrocyte end-feet processes and neurons (27).

At the BBB, glucose uptake and transport is considered to occur primarily through the glucose transporter 1 (GLUT1), also known as solute as solute carrier family 2, facilitated glucose transporter 1 (SLC2A1) (7). GLUT1 is a sodium-independent facilitated glucose transporter, displaying a substrate specificity for glucose, galactose, and dehydroascorbic acid (3, 20). Although GLUT1 is considered as the most important glucose transporter at the BBB, other glucose transporters including sodium-independent (GLUT3/SLC2A3, GLUT4/SLAC2A4) and sodium-dependent (SGLT1/SLC5A1, SGLT6/SLC5A11) have been described in vitro and in vivo (10, 23, 28, 31, 37, 39, 40, 42). However, their contribution in the uptake and diffusion of glucose across the human BBB remains to be determined.

Mutations in the SLC2A1 gene are commonly associated with GLUT1 deficiency syndrome (GLUT1DS) (33, 36). GLUT1DS is an autosomal dominant genetic disorder characterized by mutations affecting the SLC2A1 gene and impairing GLUT1 transporter activity, resulting in reduced glucose uptake at the BBB. In GLUT1DS patients, glucose cerebrospinal fluid (CSF)-to-serum concentration ratio displayed a range of 0.19 to 0.59 (16), and such a range is considered below the normal level (0.6) (30). In addition, differences in CSF glucose levels were observed between GLUT1DS patients, suggesting a possible polymorphism in GLUT1 mutations and ultimately in glucose transport phenotype. Notably, the prescription of a ketogenic diet in GLUT1DS patients, as well as in patients with refractory epilepsies, has been until now the main therapeutic approach (38). Therefore, a better understanding on how mutations in SLC2A1 genes and the contribution of other glucose transporters at the BBB may provide novel therapeutic approaches for these patients. In vitro models of the human BBB are mostly based on the hCMEC/D3 cell line (43). Yet, this cell line suffers from two major caveats: it displays poor barrier properties [transendothelial electrical resistance (TEER) < 50 Ω·cm²], resulting in their limited use for assessing drugs and nutrient permeability studies. Furthermore, such a model does not allow the modeling of neurodevelopmental disorders associated with genetic mutations.

More recently, stem cell models based on patient-derived induced pluripotent stem cells (iPSCs) have gained a momentum as a tool for modeling neurological disorders (50). iPSCs provide a patient-specific source of cells, which can be differentiated into BMECs using a differentiation protocol developed by Shusta and colleagues (18, 19). Such a protocol allows the differentiation of iPSCs into BMECs. Such cells display tight monolayers (TEER >1,000 Ω·cm²), as well as a quasisimilar gene expression profile compared with primary and immortalized human BMEC models (17, 41). Furthermore, the use of iPSCs allows the development of isogeneic models capable of differentiating astrocytes and neurons from the same lines (4, 34). Finally, the use of such differentiation protocol for disease modeling has been successfully reported to model the BBB from patients suffering from neurogenetic disorders including Allan-Herndon-Dudley Syndrome or Huntington’s disease (17, 41).

In this study, we investigated the expression profile and glucose uptake pattern in two iPSC-derived BMECs monolayers and compared such features to hCMEC/D3 monolayers, using such cell line as a referential model of the BBB.

MATERIALS AND METHODS

Cell lines.

IMR90-c4 (RRID:CVCL_C437) iPSC cell line (47) was derived from the IMR-90 somatic fibroblast cell line isolated from the lung tissue of a Caucasian female fetus and established by Nichols and colleagues (29). IMR90-c4 iPSC line was purchased from WiCell cell repository (WiCell, Madison, WI). CTR65M iPSC line (ND-41865; RRID:CVCL_Y837) was derived from fibroblasts isolated from an asymptomatic patient by Almeida and colleagues (2). This iPSC line was kindly gifted by the NINDS Human Cell and Data Repository (NHCDR) and provided by the Coriell Institute of Medical Research (Camden, NJ) and Rutgers University Cell and DNA repository (RUCDR, Rutgers, NJ). Undifferentiated iPSC colonies were maintained on human pluripotent stem cell-grade growth factor reduced Matrigel (C-Matrigel, Corning, Corning, MA) in the presence of Essential 8 medium (E8, ThermoFisher, Waltham, MA). hCMEC/D3 immortalized human brain microvascular endothelial cell line (RRID:CVCL_U985) (22, 43) was purchased from Millipore (Billerica, MA) and maintained following the manufacturer’s instructions. Cells were maintained and used for 10 passages.

BMEC differentiation.

iPSCs were differentiated into BMECs following the protocol established by Lippmann and colleagues (18, 19). iPSCs were seeded as single cells on T-Matrigel (Trevigen, Gaithersburg, MD) at a cell density of 20,000 cells/cm² in E8 supplemented with 10 μM Y-27632 (Tocris, Minneapolis, MN). Cells were maintained in E8 for 5 days before differentiation. Cells were maintained for 6 days in unconditioned medium [UM: DMEM/F-12 with 15 mM HEPES (ThermoFisher)], 20% knockout serum replacement (KOSR, ThermoFisher), 1% nonessential amino acids (ThermoFisher), 0.5% Glutamax (ThermoFisher), and 0.1 mM β-mercaptoethanol (Sigma-Aldrich). After 6 days, cells were incubated for 2 days in the presence of EC^+/+ [EC medium (ThermoFisher) supplemented with 1% platelet-poor derived serum (PDS, Alfa-Aesar, ThermoFisher), 20 ng/ml human recombinant basic fibroblast growth factor (Tocris, Abingdon, UK), and 10 μM retinoic acid (Sigma-Aldrich)]. After such maturation process, cells were dissociated by accutase (Corning) treatment and seeded as single cells on tissue culture plastic surface (TCPS) coated with a solution of collagen from human placenta (Sigma-Aldrich) and bovine plasma fibronectin (Sigma-Aldrich) (80 µg/cm² and 20 µg/cm², respectively). Twenty-four hours after seeding, cells were incubated in the presence of EC^−/− (EC medium supplemented with 1% PDS). Barrier phenotype experiments were performed 48 h after seeding.

Astrocyte and neuron differentiation.

iPSCs were differentiated into astrocytes and neurons using an adherent three-step differentiation method. Undifferentiated iPSCs were allowed to grow on C-Matrigel for 4 days before differentiation. Differentiation of these iPSCs into neural stem cells (NSCs) was induced using neural induction medium (NIM, ThermoFisher) for 11 days (46). After such induction period, NSCs were enzymatically dissociated by accutase and seeded as single cells at a cell density of 100,000 cells/cm² in the presence of 10 μM Y-27632 on C-Matrigel-coated plates. Twenty-four hours after seeding, NSCs were further differentiated into neural precursor cells (NPCs) by incubation in the presence of neural differentiation medium [NDM: human pluripotent stem cell serum-free medium (ThermoFisher), supplemented with 2% bovine serum albumin (BSA, ThermoFisher), 1% Glutamax I (ThermoFisher), 10 µg/ml human recombinant brain-derived neurotrophic growth factor (BDNF, ThermoFisher), and 10 µg/ml human recombinant glial-derived neurotrophic factor (GDNF, ThermoFisher)] for 5 days (9).

Differentiation of these NPCs into astrocytes and neurons was achieved by seeding 50,000 cells/cm² on TCPS-coated plates. Astrocytes were seeded on C-Matrigel (8 μg/cm²)-coated plates and maintained in the presence of astrocyte maturation medium (AMM: DMEM with 4.5 g/l glucose (ThermoFisher); 1% N2 supplement (ThermoFisher); and 1% fetal bovine serum (ThermoFisher)]. Neurons were seeded on poly-d-lysine (2 μg/cm², Sigma)/laminin (1 μg/cm², Sigma)-coated plates and maintained in neuron maturation medium [NMM: neurobasal medium (ThermoFisher); 2% B27 supplement (ThermoFisher) and 1% CultureOne (ThermoFisher)]. In both cases, medium was replaced every 2 days for 15 days (astrocytes) and 21 days (neurons), respectively.

Immunocytochemistry and flow cytometry.

Cells were stained on TCPS plates or dissociated with accutase and centrifuged at 200 g for 5 min (BD Biosciences) for flow cytometry experiments. Cell pellets were fixed with 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA) or 100% cold methanol (MeOH, Sigma-Aldrich). Cells were blocked for 1 h at room temperature in PBSG [PBS supplemented with 10% normal goat serum (Sigma-Aldrich)] with 0.2% Triton X-100. Cells were incubated overnight in the presence of GLUT1 (1:100; MA5-11315; ThermoFisher), GLUT3 (1:100; PA-32429; ThermoFisher), GLUT4 (1:100; PA-23059; ThermoFisher), SLC5A1 (1:100; PA5-28240), or SLC5A11 (1:100; PA5-57232; ThermoFisher) primary antibodies or in the presence of mouse (1:100; 556648; BD Biosciences) or rabbit (1:100; 550875; BD Biosciences) IgG isotype controls at same concentrations. Cells were washed with PBS containing 1% BSA (Sigma-Aldrich), and incubated in the presence of Alexa-Fluor-conjugated secondary antibodies (1:200; mouse: A32727, rabbit: A11008; ThermoFisher) for 1 h at room temperature. Cells were counterstained with DAPI (Sigma-Aldrich) and observed on a Leica DMI8 inverted epifluorescence microscope (Leica Microsystems, Wetzlar, Germany). Micrograph pictures were acquired using a ×20 long-distance (LD) Leica objective (N PLAN/0.35), with a final ×200 magnification upon acquisition. Micrograph pictures were acquired at a resolution of 694 µm × 517 µm using a Leica CDD camera coupled with Leica Acquisition Suite X (Leica). A region of interest (ROI) of 200 µm × 200 µm was randomly selected using ImageJ functions (ImageJ, NIH, Bethesda, MD). Flow cytometry samples were acquired using a FACSVerse system (BD Biosciences, San Jose, CA) and analyzed using FlowJo (FlowJo, LLC, Ashland, OR). Relative expression was obtained by subtracting the geometric mean fluorescence index (MFI) from samples versus the MFI from the IgG isotype control.

iPSC-derived BMEC barrier function.

iPSC-derived BMECs were seeded on Transwells (polyester, 0.4 μm pore size, Corning) and coated as previously described. iPSC-derived BMECs were seeded at a seeding density of 1,000,000 cells/cm², whereas hCMEC/D3 were seeded at 100,000 cells/cm². Barrier function was assessed 48 h after seeding for iPSC-derived BMEC monolayers and 6 days after seeding in hCMEC/D3 monolayers; these timepoints were determined as the most optimal for achieving the highest barrier tightness. Barrier tightness was measured by assessing both the transendothelial electrical resistance (TEER) and paracellular diffusion. TEER was measured using an EVOHM STX2 chopstick electrode (World Precision Instruments, Sarasota, FL). For each experiment, three measurements were performed for each insert and the average resistance obtained was used for the determination of the barrier function.

Glucose and mannitol permeability assays.

Prior to each experiment, TEER was assessed in each Transwell insert. Cells were rinsed with glucose-free/pyruvate-free Dulbecco’s modified Eagle’s medium supplemented with 1% PDS [DMEM(−)] for 60 min to allow the removal extracellular glucose from previous incubation medium. Nonradioactive d-glucose (Acros Organics, ThermoFisher) or nonradioactive d-mannitol (Acros Organics) was dissolved in DMEM(−). [¹⁴C-glucose] or [¹⁴C-mannitol] was added to glucose and mannitol solutions in an amount necessary to achieve a specific activity of 1 µCi/mg of nonradioactive glucose or mannitol.

In experiments involving permeability measurements, radioactive glucose or mannitol solution was added in the apical (donor) chamber whereas the basolateral (acceptor) chamber was maintained in DMEM(−) solution. Every 10 min, an aliquot of 100 µl was sampled from the basolateral chamber and replaced with nonradioactive DMEM(−) solution.

For the calculation of the clearance, the amount of glucose (or mannitol), the total activity measured at each timepoint in the basolateral chamber was converted and adjusted into amount of glucose (or mannitol) present in each timepoint. Such an amount of glucose (or mannitol) was plotted for each timepoint. Using the built-in linear regression fit algorithm (Prism, GraphPad Software, La Jolla, CA), we determined the clearance slope for each experiment, with clearance expressed as µg glucose (or mannitol)/min.

Permeability (P_e) for glucose and mannitol was calculated using the clearance slopes method developed by Perriere and colleagues (35), using the equation below:

$$1)\frac{1}{P{S}_{\mathrm{e}}}=\frac{1}{P{S}_{\mathrm{t}}}-\frac{1}{P{S}_{\mathrm{f}}}2)P_{\mathrm{e}}=\frac{P{S}_{\mathrm{e}}}{S}$$

where PS_t and PS_f indicate clearance slope of samples and blank (empty-coated) filters and S indicates insert surface area (in cm²).

Cell accumulation assays.

Cells grown on TCPS were rinse in DMEM(−) for 15 min at 37°C to remove any extracellular glucose from the previous incubation medium. Radioactive glucose and mannitol solutions were prepared as described in the previous section. Cells were incubated for 60 min at 37°C. Following the incubation period, cells were briefly washed three times with ice-cold PBS and homogenized with 100 µl PBS containing 0.2% Triton X-100 (Sigma-Aldrich) for 10 min. Ninety microliters of the cell lysates were used for the determination of total radioactivity, whereas 10 µl of the cell lysates were used for protein concentration assays. Sample radioactivity was assessed by adding 90 µl samples with 5 ml ScintiSafe 30% liquid scintillation cocktail (Fisher Chemicals, ThermoFisher) and counted using a Beckman-Coulter LS6500 liquid scintillation counter (Beckman-Coulter, Pasadena, CA). Protein concentrations for each cell lysate were determined by BCA protein quantitation assay (Pierce, ThermoFisher). For each sample, 5 µl of cell lysates were incubated with 200 µl of BCA reagent for 30 min at 37°C. Following incubation, sample absorbance was determined at 540 nm using a Synergy MX2 plate reader (Bio-Tek, Burlington, VT). Protein concentration was determined using bovine serum albumin as protein standard.

Inhibition experiments.

Cells were incubated in the presence of DMEM(−) supplemented with 10 µM glucose transporter inhibitor II (GTI, Calbiochem), 10 µM phloridzin (PHZ, Tocris Biomedical), 10 µM phloretin (PHT, Tocris), or 10 µM STF31 (Tocris) for 1 h prior to experiments. At the beginning of experiments, glucose solution (1 mg/ml, 1 µCi/ml) supplemented with inhibitor was added in the apical (donor) chamber whereas DMEM(−) supplemented with inhibitors was added in the bottom chamber. Cells incubated in the presence of 0.1% DMSO (vehicle) served as controls. Cell accumulation and cell permeability assays were performed as previously.

Statistics.

Cells were randomly assigned treatment conditions prior to each experiment. Data are represented as means ± SD from three independent experiments. One-way analysis of variance (ANOVA) coupled with Dunnett (or Kruskal-Wallis) test analyses were performed using Prism 7.0 built-in package (GraphPad Software). A P value equal or less than 0.05 indicated a statistically significant difference between one or more groups.

RESULTS

iPSC-derived BMECs display a glucose transporter expression profile similar to hCMEC/D3 cells.

Several glucose transporters have been described in the literature as expressed at the blood-brain barrier in various in vitro and in vivo models (10, 23, 28, 31, 37, 39, 40, 42). Therefore, we first investigated the expression profile of several glucose transporters at the BBB in hCMEC/D3 and in our iPSC-derived BMECs by immunofluorescence (Fig. 1A).

Fig. 1.

Expression profile of glucose transporters in brain microvascular endothelial cells (BMECs). A: representative expression of glucose transporter 1 (GLUT1), GLUT3, GLUT4, sodium-dependent glucose transporter 1 (SGLT1), and SGLT6 in IMR90-derived BMECs (IMR90), CTR65M-derived BMECs, and hCMEC/D3 monolayers. Cells were fixed in 100% cold methanol (GLUT1, GLUT3, GLUT4) or 4% paraformaldehyde (SGLT1, SGLT6) before staining. Exposure time was adjusted to negative control (IgG isotype, secondary antibody, data not shown). Note the dotted immunostaining profile for SGLT6, suggesting a nuclear colocalization. Scale bar, 50 µm. B and C: expression profile of glucose transporters in induced pluripotent stem cells (iPSC)-derived astrocytes (B) and neurons (C). Staining protocol followed the same procedure as BMECs. DAPI (blue) served as a nuclei counterstain. Scale bar, 50 µm. D: representative flow cytometry histogram profile for GLUT1, GLUT3, GLUT4, SGLT1, and SGLT6 in iPSC-derived BMECs and hCMEC/D3. Histograms depict the relative distribution of cells based on fluorescence intensity (x-axis) and by the relative frequencies of such events (y-axis). Red histograms depict cells incubated in the presence of IgG isotype controls, and blue histograms depict cells incubated in the presence of the protein of interest. E: mean fluorescence index (MFI) of the different glucose transporters analyzed by flow cytometry. MFIs were obtained by calculation of the geometric means in both samples and IgG isotype controls. Plotted MFI values were obtained after subtraction of the IgG MFI from the sample MFI. ND, not determined (<10% cell population were displaying a positive expression). *P < 0.05 compared with hCMEC/D3 cells; n = 4/group.

In our hands, all three BMEC monolayers showed positive expression for GLUT1, GLUT3, and GLUT4, such expression being in agreement with the literature (23, 28).

Interestingly, none of our BMECs showed a positive expression for SLGT1 despite the documentation of SGLT1 at mRNA levels in hCMEC/D3 cells and in vivo (23, 31, 40). We noted a positive expression of SGLT6 in all three BMEC monolayers; however, such transporters appeared localized in the nuclear region, contrasting with the perimembraneous localization of the three GLUTs tested.

To contrast such an expression profile observed in iPSC-derived BMECs with other cells of the central nervous system, we differentiated astrocytes and neurons from the same iPSC lines and assessed the expression profile of the same transporters in these cells (Fig. 1, B and C). iPSC-derived astrocytes showed a relatively high GLUT1 expression in both IMR90 and CTR65M lines, as well as a positive but faint staining for GLUT3. Among all three cell types investigated in this study, neurons showed the less diverse expression of glucose transporters. Neurons showed a weak immunostaining for GLUT3 and GLUT4. Interestingly, we noted a higher expression of SLGT6 in CTR65M-derived astrocytes and neurons compared with IMR90-derived cells.

Next, we quantified the relative expression of these glucose transporters in BMECs using flow cytometry (Fig. 1, D and E). We noted notable differences in the relative expression (as expressed as geometric mean fluorescence intensity, MFI) between iPSC-derived BMECs and hCMEC/D3 cell monolayers. As expected (34), both iPSC-derived BMECs showed higher GLUT1 expression (1.5-fold) than hCMEC/D3. Notably, GLUT3 protein expression was higher in iPSC-derived BMECs compared with hCMEC/D3 cells; whereas GLUT4 and SGLT6 were lower in iPSC-derived BMECs than hCMEC/D3 cells. Taken together, iPSC-derived BMECs showed a glucose transporter expression profile similar to hCMEC/D3; however, differences in their expression levels were observed between iPSC-derived BMECs and hCMEC/D3.

BMECs display a selective uptake for glucose over mannitol.

As glucose uptake and permeability across the blood-brain barrier occurs via a facilitated transport via glucose transporters, it is important to demonstrate that such transport across iPSC-derived BMECs is solely mediated by a transcellular route.

To exclude the presence of a paracellular (through cell junctions) route in the glucose diffusion across iPSC-derived monolayers, we assessed the uptake of radiolabeled d-glucose in our cells (Fig. 2A). Both iPSC lines showed a glucose uptake that was 10 times higher than hCMEC/D3 cells. In the opposite, the uptake of d-mannitol (a monosaccharide similar to glucose but not uptaken at the BBB) was negligible and similar to hCMEC/D3 cells.

Fig. 2.

Glucose uptake and diffusion across brain microvascular endothelial cell (BMEC) monolayers involve a specific mechanism. A: mannitol and glucose uptake profile in BMECs grown on a tissue culture plastic surface (TCPS). Cells were incubated for 60 min in the presence of 1 g/l mannitol or glucose (total activity: 1 µCi/ml) prior to washing with ice-cold PBS and cell lysis. Note the residual values obtained with mannitol compared with glucose, confirming the presence of a specific uptake mechanism. B: transendothelial electrical resistance (TEER) values of the induced pluripotent stem cell (iPSC)-derived BMECs used for glucose transport experiments. Note the low TEER values obtained in hCMEC/D3 indicative of a poor barrier tightness. C: comparative permeability profiles of glucose and mannitol in iPSC-derived BMECs and hCMEC/D3 cells. Cells were incubated with 1 g/l mannitol or glucose (total activity: 1 µCi/ml) in the apical chamber. Note the low permeability (P_e) values obtained with mannitol (paracellular marker) obtained in iPSC-derived BMECs compared with hCMEC/D3, indicative of a better barrier tightness. In contrast, glucose diffusion across iPSC-derived BMECs was 5 times faster than mannitol, indicative of a facilitated transport mechanism. **P < 0.01 in comparison between glucose and mannitol; n = 4/group. D: regression analysis between TEER and glucose clearance. Prior to glucose diffusion experiments, TEER in each insert was measured and plotted against the glucose clearance slope obtained from permeability assays. Note the absence of correlation between TEER and glucose clearance for iPSC-derived BMECs, whereas an inverse correlative trend can be observed for hCMEC/D3 cells.

Next, we compared the barrier tightness of iPSC-derived BMECs to hCMEC/D3 cells (Fig. 2B). As expected (34), both iPSC lines showed a barrier function that was 20 times tighter than hCMEC/D3 cells, suggesting a presence of tight monolayers.

Finally, we compared the diffusion profile of glucose across our monolayers and compared with mannitol (paracellular diffusion only). iPSC-derived BMECs showed mannitol P_e values ~3 × 10⁻⁴ cm/min; such values were slightly higher than the apparent permeability value (noncorrected permeability value) reported by Helms and colleagues (14) in primary porcine BMEC monolayers but less than half the P_e value than reported in hCMEC/D3 cells.

Using the P_e values to directly compare the permeability of the different BMEC monolayers to glucose and mannitol (Fig. 2C), the permeability of glucose was 5 times higher than mannitol in both iPSC lines. In contrast, hCMEC/D3 monolayers failed to show a difference in permeability between glucose and mannitol. Such failure reflects the inability of hCMEC/D3 cells to form tight monolayers capable to discriminate paracellular fluxes from transcellular fluxes. In summary, our data suggest that iPSC-derived BMECs appear more suitable for glucose permeability studies than hCMEC/D3 cells.

Glucose uptake in iPSC-derived BMECs involves a GLUT-dependent process.

Using the accumulation assay, we investigated the kinetic profile of glucose uptake in BMEC monolayers (Fig. 3). We first assessed the presence of a saturable glucose uptake in iPSC-derived BMECs and compared it to hCMEC/D3 cells (Fig. 3A). All three monolayers showed a Michaelis-Menten kinetic profile following uptake at different glucose concentrations tested. Experimental K_m values obtained were 0.63 (3.5 mM), 0.31 (1.7 mM), and 0.39 g/l (2.1 mM) for hCMEC/D3, IMR90, and CTR65M BMEC monolayers, respectively. Such K_m values were similar to GLUT1 and GLUT3 K_m values as reported in the literature (12).

Fig. 3.

Glucose uptake in induced pluripotent stem cell (iPSC)-derived brain microvascular endothelial cells (BMECs) involves a saturable and glucose transporter (GLUT)-dependent mechanism. A: glucose uptake assay in iPSC-derived BMECs and hCMEC/D3 cells. Cells were incubated for 60 min with various glucose concentrations (specific activity of 1 µCi/mg) before cell lysis. The amount of intracellular glucose was normalized to the total amount of proteins in cell lysates. B: uptake inhibition studies. IMR90- and CTR65M-derived BMECs were incubated in the presence of 10 µM glucose transporter inhibitor II (GTI), phloridzin (PHZ), phloretin (PHT), or STF31. Cells preincubated in the presence of such inhibitors in glucose-free/pyruvate-free DMEM [DMEM(−)] for 1 h; such inhibition was maintained throughout the experiment. Cells were allowed to incubate in the presence of glucose solution (1 g/l concentration, 1 µCi/ml) for another hour prior to cell lysis. Control wells were incubated in the presence of vehicle (0.1% DMSO). *P < 0.05 and **P < 0.01 vs. control group; n = 4.

Next, we assessed the ability to inhibit such glucose uptake with pharmacological inhibitors (Fig. 3B). Treatment with 10 µM phloridzin (PHZ, an SGLT inhibitor) resulted in a mild decrease in glucose uptake in both iPSCs compared with control group, whereas treatment with 10 μM phloretin (PHT, the aglycone form of PHZ and GLUT inhibitor) or glucose transporter inhibitor II (GTI) resulted in a significant decrease in glucose uptake in both iPSC lines, suggesting that glucose uptake was mostly driven through a GLUT-dependent mechanism. Treatment with STF31 (a GLUT1 inhibitor) resulted in a significant decrease in glucose uptake; however, such decrease was less accentuated than other GLUT inhibitors, suggesting the contribution of other GLUTs in glucose uptake in such cells. Taken together, our data suggest the presence of a saturable glucose uptake in iPSC-derived BMECs; such uptake appears to involve one or several GLUT isoforms.

Glucose diffusion across the blood-brain barrier occurs via a saturable, GLUT-dependent mechanism.

As we have previously mentioned, glucose transport across the BBB occurs through a transcellular mechanism. Therefore, to demonstrate the relevance of iPSC-derived BMECs as an in vitro model for permeability studies, it is important to demonstrate the ability of glucose to cross the BBB and reach the basolateral compartment.

Using Transwell inserts, we measured the glucose permeability profile in our iPSC-derived BMECs and compared it to hCMEC/D3 monolayers (Fig. 4A). As previously observed (Fig. 3A), iPSC-derived BMECs showed a higher permeability value for glucose than hCMEC/D3 for any concentrations tested. With the exception of hCMEC/D3 monolayers (in which higher TEER values were inversely correlated with decreased glucose clearance), the impact of barrier tightness (as measured in TEER) had no effect on glucose permeability (as measured by clearance) in iPSC-derived BMECs (Fig. 2D). In all three BMECs, we observed a Michaelis-Menten kinetics reflecting the presence of a saturable diffusion profile. Estimation of the K_m values in our permeability studies showed very similar values compared with those obtained in uptake studies (Fig. 3A), with K_m values of 3.54 mM, 1.76 mM, and 2.18 mM in hCMEC/D3, IMR90-, and CTR65M-derived BMECs, respectively.

Fig. 4.

Glucose transport across induced pluripotent stem cell (iPSC)-derived brain microvascular endothelial cells (BMECs) involves a saturable and glucose transporter (GLUT)-dependent mechanism. A: glucose permeability assay in iPSC-derived BMECs and hCMEC/D3 cells. Cells were incubated in the presence of various glucose concentrations (specific activity of 1 µCi/mg) in the apical chamber, whereas the basolateral chamber was maintained in DMEM(−). Aliquots from the basolateral chamber were obtained at regular time intervals. B: diffusion inhibition studies. IMR90- and CTR65M-derived BMECs were incubated in the presence of 10 µM glucose transporter inhibitor II (GTI), phloridzin (PHZ), phloretin (PHT), or STF31 in both the apical and basolateral chambers. Cells preincubated in the presence of such inhibitors in DMEM(−) for 1 h; such inhibition was maintained throughout the experiment. After such incubation time, cell medium in the apical chamber was replaced with glucose solution (1 g/l concentration, 1 µCi/ml). Sampling occurred as previously mentioned. Control inserts were incubated in the presence of vehicle (0.1% DMSO). *P < 0.05 and **P < 0.01 vs. control group; n = 4.

Next, we assessed the effects of pharmacological inhibitors on glucose diffusion across the monolayer (Fig. 4B). Treatment with PHZ modestly reduced glucose diffusion in iPSC-derived BMECs. Interestingly, the inhibitory effect of PHZ was more accentuated in CTR65M-derived BMECs compared with IMR90-derived BMECs. As expected, PHT showed a significant decrease in glucose clearance across the monolayers; however, such inhibition was less marked than observed in the uptake assay (~30% inhibition in clearance vs. ~50% inhibition in uptake). Maximal inhibition in iPSC-derived BMECs was observed in the presence of GTI, whereas GLUT1 inhibition by STF31 resulted in a modest yet significant decrease in glucose clearance compared with control. Again, such inhibition was more accentuated in CTR65M monolayers compared with IMR90. In conclusion, glucose diffusion across the iPSC-derived BMEC monolayers involves a saturable process and showed similar contribution of GLUTs compared with that observed in uptake assays.

DISCUSSION

Glucose uptake and diffusion across the blood-brain barrier (BBB) constitutes an important physiological aspect in brain glucose metabolism, as glucose represents the major source of energy to the brain.

There is a growing literature demonstrating the contribution of impaired glucose metabolism at the BBB in several neurological diseases including Alzheimer’s disease (AD) and GLUT1 deficiency syndrome (GLUT1DS) (37, 45). Yet, in vivo models provide a very limited understanding of the pathophysiological processes undergoing at the BBB in such situations and raise the need of suitable in vitro models. In this study, we investigated the expression and activity of glucose transporters in iPSC-derived BMEC monolayers and compared with hCMEC/D3 cells, an immortalized human BMEC line commonly used as a reference for modeling the human BBB in vitro (13).

In our hands, iPSC-derived BMECs derived from the two independent iPSC cell lines showed a GLUT expression profile similar to that of hCMEC/D3 cells. Such observation nicely correlates with two recent publications that showed a very high similarity in gene expression profiles between BMECs generated from iPSC lines to primary human BMECs and freshly isolated cells (17, 41). Such comparative transcriptomics obtained from independent iPSC lines using the same differentiation protocol as used by our group (18, 19) clearly suggest the relevance of stem cell-derived models for modeling the BBB in vitro.

Although we did not extensively characterize the expression profile of glucose transporters in iPSC-derived astrocytes and neurons, we did observe a distinct glucose transporter expression pattern between astrocyte and neurons. GLUT1 was the major glucose transporter expressed in iPSC-derived astrocytes, whereas GLUT3 isoform was the predominant form expressed in iPSC-derived neurons. These expression patterns were very similar to the expression pattern described in the literature using in vitro (21) and in vivo (48, 49) models.

Notably, we failed to observe the expression of SGLT1 at protein levels in our iPSC-derived BMECs and hCMEC/D3 monolayers, contradicting the existing literature. Expression of SGLT1 has been reported in primary rat and bovine BMECs (10, 28, 42). The presence of SGLT-mediated glucose transport across BMECs cannot be excluded, however, as we still noted a moderate inhibition in glucose uptake and diffusion in the presence of phloridzin (an SGLT inhibitor). We cannot exclude that SGLT1 (and potentially other SGLTs) expression at the BBB may occur only following injury. In particular, SGLT1 expression in primary bovine BMECs was significantly upregulated following oxygen-glucose deprivation stress (42). We are currently investigating the possible difference in expression of such transporters following stroke injury (using oxygen-glucose deprivation stress) in our model, as such cells displayed a response to cerebral hypoxia/ischemia similar to in vivo (32).

Among the different glucose transporters investigated in this study, SGLT6 expression pattern was unique. Unlike the expression of other glucose transporters (cytoplasmic and cell membranes), SGLT6 immunostaining displayed a perinuclear pattern. Interestingly, the main substrate for SGLT6 is myo-isonitol (MI, also known as isotinol). MI is considered as one of the most abundant metabolites in the central nervous system with a contribution in several neurological disorders, with brain concentration exceeding plasma concentration (11). Yet, the transport of MI across the BBB remains unclear. Therefore, our model can help elucidate the presence of a possible MI transport across the BBB.

In this study, we focused on the expression of a limited number of glucose transporters described in the literature. However, we cannot exclude the expression of other glucose transporters. Therefore, a comparative transcriptomic analysis between iPSC-derived BMECs and human BMECs freshly isolated from brain microvessels can provide us with potential hindsight on novel glucose transporters.

In addition to the expression profile, we compared the uptake and diffusion profile of glucose in iPSC-derived BMECs compared with hCMEC/D3 cells. iPSC-derived BMECs show interesting barrier properties on par with bovine and porcine primary BMEC monolayers (considered as the best primary cultures in terms of barrier tightness for drug permeability studies) (13). Yet, studies providing a direct comparison in terms of drug permeability between iPSC-derived models and existing in vivo and in vitro models remain unclear. Our study suggests that iPSC-derived BMECs outperform hCMEC/D3 in terms of glucose transport, as we noted uptake and diffusion values five times higher in iPSC-derived BMECs compared with hCMEC/D3 cells. Such difference cannot be solely explained by differences in GLUTs expression, as we noted only a twofold difference. We speculate that the discrimination between paracellular and transcellular fluxes may be a determining factor for drug and nutrient permeability studies.

Although we noted similar behavior between uptake and diffusion assays, we noted some differences between these two techniques. Hence, we cannot exclude differences in expression of glucose transporters between cells grown on TCPS and cells grown on Transwell inserts. Thus, a comparative expression analysis of glucose transporters between cell grown on TCPS versus cells grown on Transwells will help us shed some light into these observations.

Pharmacological inhibition experiments performed in this study suggest the presence of one or several GLUT transporter uptake and diffusion of glucose across the BBB. However, a major limitation we encountered is the absence of selective GLUT isoform inhibitors (with the exception of STF31). Therefore, we are currently investigating the ability to confirm our results by selectively silencing one or several GLUTs using shRNA and gene-editing (CRISPR/Cas9) techniques. Such an approach will allow us to confirm our observations and also highlight an interplay between the different GLUT isoforms that may result in some compensatory mechanism.

A limitation of our model is marked by the absence of cocultures with astrocytes and/or neurons. Astrocytes (and to a lesser extent neurons) are well-known to upregulate BMEC barrier tightness upon cocultures (1, 4, 5, 8, 15, 25, 26, 34, 44). We cannot exclude that the presence of astrocytes may upregulate endothelial cell polarity and the localization of glucose transporters at the apical and/or basolateral sides. In addition, coculture models can provide an integrative model potentially suitable to investigate endothelial-astrocyte coupling in terms of glucose and lactate metabolism. We are currently investigating changes in glucose permeability following cocultures.

In conclusion, iPSC-derived BMECs appear to show interesting potential for modeling glucose uptake and diffusion across the BBB. Such a model can provide a useful investigative tool to model GLUT1DS in a dish, similarly to other disease-specific models of BBB from patients suffering from Allan-Herndon-Dudley Syndrome or Huntington’s disease (17, 41).

GRANTS

This study was funded through a GLUT1 deficiency foundation research grant and institutional funds (to A. J. Al-Ahmad).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author.

AUTHOR CONTRIBUTIONS

A.J.A.-A. conceived and designed research; performed experiments; analyzed data; interpreted results of experiments; prepared figures; drafted manuscript; edited and revised manuscript; approved final version of manuscript.