Cell Culture
CiPTEC-OAT1 were obtained as described previously (9,10). Briefly, cells were retrieved from urine from a healthy volunteer in compliance with the guidelines of the Radboud Institutional Review Board and conditionally immortalized via transduction with the temperature-sensitive mutant of SV large T antigen (SV40T) and human telomerase reverse transcriptase (hTERT) (9). Transduction of OAT1 in ciPTEC was performed using lentiviral particles containing genes encoding for human OAT1 (10).
Cells were sub-cultured at a dilution of 1:2 to 1:6 at 33°C and 5% (v/v) CO2 to allow proliferation and were used from passage numbers 48 to 60 for experiments. Cells were cultured in a 1:1 mixture of Dulbecco’s modified Eagle’s medium (DMEM) and Ham’s F-12 nutrient mixture without phenol red (DMEM HAM’s F12, Life Technologies, Paisley, UK) and supplemented with insulin (5 μg/ml, Sigma-Aldrich, Zwijndrecht, the Netherlands); transferrin (5 μg/ml, Sigma-Aldrich); selenium (5 ng/ml, Sigma-Aldrich); hydrocortisone (36 ng/ml, Sigma-Aldrich); epidermal growth factor (10 ng/ml, Sigma-Aldrich); tri-iodothyrine (40 pg/ml, Sigma-Aldrich); 1% (v/v) penicillin/streptomycin (Life Technologies); and 10% (v/v) fetal calf serum (FCS, Greiner Bio One, Kremsmuenster, Austria), further referred to as PTEC complete medium.
For 2D static cell culture, cells were seeded at 55,000 cells/cm2 in 6-well plates and cultured for 1 day at 33°C 5% (v/v) CO2 without penicillin/streptomycin to allow proliferation. Then, cells were cultured for 7 days at 37°C 5% (v/v) CO2 to induce maturation into confluent cell monolayers. Cells were harvested on day 8 after seeding.
For 3D FSS cell culture, a 400 μm two-lane OrganoPlate (Mimetas, Leiden, the Netherlands) was used for culturing ciPTEC-OAT1 in a microfluidic platform. Prior to seeding, 50 μl of Hank’s balanced salt solution (HBSS, Life Technologies) was added to the observation window (Fig. 1a). An ECM mix containing collagen I (4 mg/ml, AMSbio Cultrex 3D Collagen I Rat Tail); sodium bicarbonate (3.7 mg/ml, Sigma-Aldrich); and HEPES (100 mM, Life Technologies) was prepared and 2 μl was added to the gel channel (17). Then, the OrganoPlate was placed at 37°C 5% (v/v) CO2 for 30 min to allow polymerization of the ECM. Proliferating cells were harvested from a culture flask using accutase (Sigma-Aldrich), centrifuged (300×g, 5 min) and then resuspended in PTEC complete medium. Per chip, 40,000 cells in a suspension of 2 μl were seeded in the medium-channel inlet (Fig. 1a, b). In chips used for background correction of fluorescent signal, 2 μl of medium was used. The OrganoPlate was then placed vertically at 33°C 5% (v/v) CO2 for 30 min with the gel channel facing down to allow cells to attach to the ECM (Fig. 1b). Next, 30 μl of PTEC complete medium was added to the medium channel inlet. The OrganoPlate was oriented horizontally again at 33°C 5% (v/v) CO2 for 4 to 6 h. Upon addition of another 30-μl PTEC complete medium to the medium channel inlet, 60-μl PTEC complete medium was added to the medium channel outlet. CiPTEC-OAT1 in the OrganoPlate was cultured at 33°C 5% (v/v) CO2 to allow proliferation without any FSS for 3 days. Medium was supplemented with 1% (v/v) penicillin/streptomycin when culturing at 33°C.
At day 3 after seeding, medium channels were refreshed with PTEC complete medium without penicillin/streptomycin and the OrganoPlate was placed on an interval rocker platform (angled at − 7° or + 7° from base, 8-min interval) at 37°C 5% (v/v) CO2 to allow the cells to maturate under FSS induced by a bi-directional flow (Fig. 1b, c). Experiments in the OrganoPlate were performed on day 8 after seeding, unless stated otherwise. Medium was refreshed every 2 to 3 days.
For long-term cell culture of ciPTEC-OAT1 in the OrganoPlate, chips were pre-coated (30 min, 37°C) prior to seeding with L-3,4-dihydroxydiphenylalanine (L-DOPA, 2 mg/ml in 10 mM Tris pH 8.5, Sigma-Aldrich) as earlier described (18). Cells were then cultured at 33°C on an interval rocker, as above described, for 20 days.
Immunofluorescence staining of ciPTEC-OAT1 in the OrganoPlate
At days 8 and 20 after seeding, PTEC complete medium was aspirated from the medium channel and cells were fixated with 3.7% (w/v) formaldehyde (Sigma-Aldrich) in HBSS. The plate was incubated for 15 min at room temperature. For all described incubation and wash steps, 50 μl was added to the inlet and outlet of the medium channel, and plate was placed at angle to allow perfusion, unless stated otherwise. Fixative was aspirated and medium channels were washed twice with washing solution (4% (v/v) FCS in HBSS). Then, cells were permeabilized by adding 0.3% (v/v) Triton-X100 (Sigma-Aldrich) in HBSS and incubated for 10 min at room temperature. Cells were subsequently washed with washing solution and then incubated with blocking solution (2% (v/v) FCS, 2% (w/v) bovine serum albumin (BSA, Sigma-Aldrich), 0.1% (v/v) Tween-20 (Sigma-Aldrich) in HBSS) for 45 min at room temperature. Primary and secondary antibodies were diluted in blocking solution and perfusion during incubation was allowed by adding 30 μl to the medium channel inlet and 10 μl to the outlet. Primary antibodies against zonula occludens-1 (ZO-1, 1:125, rabbit, Thermo Fisher), a tight-junction protein; acetylated-tubulin (clone 6-11B-1, 1:1000, mouse, Sigma-Aldrich), a constituent of the primary cilium; and pericentrin (1:500, rabbit, Abcam), a centrosome marker, were used and incubated for 60 min at room temperature. Next, cells were washed three times with washing solution and then incubated with secondary antibodies Alexa Fluor® 488 goat anti-rabbit (1:250, Life Technologies or Abcam), Alexa Fluor® 488 goat anti-mouse (1:250, Life Technologies), and Alexa Fluor® 647 goat anti-rabbit (1:250, Life Technologies) for 30 min at room temperature. Cells were then washed three times, followed by addition of probes in the last wash step to stain F-actin and nuclei with ActinRed 555 (two drops per milliliter in HBSS, Life Technologies) and DRAQ5 (1:1000 in HBSS, Abcam) or DAPI (300 nM in HBSS, Life Technologies), respectively. Fluorescent images of the ZO-1 staining were acquired with a Leica TCS SP5 laser point confocal microscope (Leica, Wetzlar, Germany) or a Zeiss LSM880 confocal scanning microscope (Carl Zeiss, Jena, Germany) using a ×20 objective. Z stacks were acquired using a 2 μm interval over a 200–220 μm range. To investigate polarization of the PTEC monolayer of ciPTEC-OAT1 in the OrganoPlate, fluorescent images were acquired of acetylated-tubulin and pericentrin staining with a Zeiss LSM880 confocal scanning microscope using a ×40 water immersion objective. A range of 45 μm was imaged using a 0.88-μm interval between Z stacks. Images were adjusted for windows and level, and then 3D images were reconstructed in Fiji (version 1.51n) (19).
Gene Expression of ciPTEC-OAT1
Cells were harvested from 6-well plates using Accutase, and then RNA was extracted using the RNeasy Mini kit (QIAGEN, Venlo, the Netherlands). RLT lysis buffer was added to the cell pellet and RNA was then extracted according to the protocol provided by the manufacturer.
RNA from cells in the OrganoPlate was isolated using the RNeasy Micro kit (QIAGEN). From each chip, RNA was isolated by adding 50 and 25 μl RLT lysis buffer to the medium channel inlet and outlet, respectively. Cell lysates of two chips were then pooled and RLT lysis buffer was added up to a volume of 350 μl. The extracted RNA was purified following the manufacturer’s protocol.
M-MLV reverse transcriptase (Promega, Madison, USA) was used to synthesize complementary DNA (cDNA) according the manufacturer’s protocol. Human kidney cortex material was used as reference. Tissues unsuitable for transplantation were obtained for research purposes from three donors, who gave permission and which was registered in the Dutch Donor registry. RNA isolation and cDNA synthesis were performed as previously described (20). The expression levels of mRNA were quantified using gene-specific primer-probe sets from Applied Biosystems (Bleiswijk, the Netherlands): GAPDH (hs99999905_m1), HPRT1 (hs02800695_m1), SLC22A6 (hs00537914_m1), ABCB (hs01067802_m1), ABCC2 (hs00166123_m1), ABCC4 (hs00195260_m1), SLC22A2 (hs01010723_m1), and TaqMan Universal polymerase chain reaction (PCR) Master Mix. Quantitative PCR reactions were carried out using CFX96-Touch Real-Time PCR System (BioRad, Veenendaal, the Netherlands) and analyzed with BioRad CFX Manager (version 3.1). Expression levels of GAPDH were used as reference gene.
3D Drug Efflux Assays in the OrganoPlate
Functionality of P-gp and MRP2/4 was assessed based upon accumulation of substrates, as previously described for ciPTEC in 2D (Fig. 1c) (12). Prior to measuring efflux activity of P-gp and MRP2/4 in the OrganoPlate, medium-channel inlet and outlets were washed with 50 μl KHH at 37°C. Medium channels were then perfused two times. Then, cells were exposed to substrates with or without inhibitors or corresponding vehicle for 1 h at 37°C.
For P-gp activity assessments, calcein-AM (2 μM, Life Technologies) was used as substrate and PSC833 (10 μM, Tocris, Bristol, UK) was used as inhibitor. Calcein-AM and PSC833 were dissolved in DMSO (Sigma-Aldrich). All dilutions were prepared in Krebs-Henseleit buffer (Sigma-Aldrich) supplemented with HEPES (10 mM, Sigma-Aldrich) and adjusted to pH 7.4 (KHH).
Chloromethylfluorescein-diacetate (CMFDA, 1.25 μM, Molecular Probes) was used to study MRP2/4 activity. A combination of PSC833 (10 μM); KO143 (10 μM, Sigma-Aldrich); and MK571 (10 μM, Sigma-Aldrich) was used to inhibit efflux of CMFDA and glutathione-methylfluorescein (GS-MF) via P-gp, BCRP, and MRP2/4, respectively. CMFDA and KO143 were dissolved in DMSO, while MK571 was dissolved in Milli-Q water, requiring appropriate vehicle controls. All dilutions were prepared in KHH.
After substrate incubations, cell nuclei were stained with Hoechst33342 by first washing with KHH, followed by staining with Hoechst33342 (10 μg/ml, Molecular Probes) for 10 min at room temperature. Cells were co-incubated with PSC833 (10 μM), KO143 (10 μM), and MK571 (10 μM) during staining with Hoechst33342 to inhibit all efflux transport processes. Subsequently, cells were washed and efflux of calcein-AM and GS-MF was arrested by adding ice-cold KHH to medium channel. FSS in medium channel during experiment and washing was assured by adding 80 μl to the medium channel inlet and 20 μl to the medium channel outlet.
Fluorescent Imaging in the OrganoPlate
Fluorescence of intracellular accumulation of calcein or GS-MF and Hoechst33342 in ciPTEC-OAT1 cultured in the OrganoPlate were imaged in situ using a Becton Dickinson (BD) Pathway 855 high-throughput microscope (BD Bioscience, Breda, The Netherlands) using a ×10 objective (Fig. 1c). The BD Pathway 855 microscope was compatible with the two-lane OrganoPlate and allowed to acquire images of the observation window of each chip using two different fluorescent dyes with wild-field microscopy and spinning disk confocal microscopy at room temperature. A macro was developed to acquire fluorescent and bright-field images using BD AttoVision (version 1.6, BD Bioscience). Z height was determined per experiment by determining the top and bottom of multiple chips in the OrganoPlate visually. The average height was used as starting point for acquiring five Z planes with a 10 μm interval, covering 50 μm in Z direction, generating focused images of cells cultured against ECM. First, calcein and GS-MF fluorescence images were acquired (488-nm excitation and 520-nm emission filters) with exposure time and gain set at 0.5 s and 100, respectively. Hoechst33342 images were acquired subsequently (360-nm excitation filter and 435-nm emission filter), with exposure time and gain set at 0.01 s and 100, respectively. Afterwards, bright-field images of each chip were acquired using an exposure time of 0.02 s.
Image Processing
An image analysis method (Fig. 1c) was set up to analyze the full data set using custom made macros in Fiji. Initially, the region of interest (ROI) was selected from bright-field images in which the gel channel, medium channel and the phase guide were clearly visible, by focusing at the interface of ECM and medium. The same area size was used for all images. All raw fluorescence images (uncompressed 12-bit TIFF) were cropped based upon the interface selection and converted from a pixel to micrometer scale. Then, the raw fluorescence images were corrected for background noise by subtracting images of chips with ECM without any cells exposed to either calcein-AM or CMFDA. Background-corrected images were saved to an uncompressed 16-bit TIFF file and the average intensity was calculated (see macro in Supplementary Material). In parallel, a mask was created for each image using 17 threshold algorithms available in Fiji (Table S1 in Supplementary Material) to automatically determine localization of intracellular accumulated fluorescent calcein or GS-MF. Upon running a median image filter (radius size five pixels), a cell selection mask was created via the selected algorithm to create a binary image. Pixels from the edges of the cell selection mask in the binary image were removed using “erode” in Fiji. The binary image represented a cell selection and was used to analyze fluorescence intensity of each Z stack per chip.
As part of the image quality control (QC), only in-focus images of cells cultured at the ECM and medium interface were selected and reviewed by three independent observers (Fig. 1c). An image was defined as out-of-focus if no in-focus cell in the ROI were scored by at least two of the three observers.
Manual interventions, such as determination of out-of-focus images are not compatible with HTS. Therefore, selected image QC metrics were established to automatically exclude out-of-focus images: focus score, image correlation, and power log-log slope (PLLS), which all have been described in literature for image analysis (21,22,23,24). An image QC method was set up using these image QC metrics in CellProfiler (version 2.2.0) (25,26). For the focus score, the deviation of pixel intensity was assessed upon normalization per image. The correlation of the grayscale value of neighboring pixel was evaluated for image correlation. Pixel size for image correlation and focus score was empirically based upon the area of the observed intracellular accumulation of calcein or GS-MF and set at 20, 50, 100, 200, and 500 pixels. For the PLLS, the slope of power spectral density versus the spatial frequency, both on a log scale, was plotted. All images were automatically binned based upon the image QC metrics in GraphPad Prism (version 5.03). Multiple cutoff values were compared per image QC metric. The efficiency of the automated image QC metrics was compared to visually reviewing of out-of-focus scoring as described earlier, which was defined as ground truth. For all image QC metrics, the precision and recall rates were determined, as well as the harmonic mean of the precision and recall rates, also known as F score. Precision, recall, and F score were calculated as follows:
$$ F\ \mathrm{score}=\frac{2\times \mathrm{precision}\times \mathrm{recall}}{\mathrm{precision}+\mathrm{recall}} $$
where precision is:
$$ \mathrm{precision}=\frac{\mathrm{true}\ \mathrm{positives}}{\mathrm{true}\ \mathrm{positives}+\mathrm{false}\ \mathrm{positives}} $$
and recall is:
$$ \mathrm{recall}=\frac{\mathrm{true}\ \mathrm{positives}}{\mathrm{true}\ \mathrm{positives}+\mathrm{false}\ \mathrm{negatives}} $$
Average intensities of visual observation in-focus images were calculated per chip (Fig. 1c). Two to eight chips were analyzed per condition and repeated in at least three independent experiments. All groups were checked for outliers using Grubbs’ test for outliers (using a significance level of α = 0.05) and finally normalized to average fluorescence intensity in vehicle condition per experiment.
The Z′ factor was used to index the thresholding algorithms and overall intensity per image and condition to assess quality and optimize the data analysis methods. The Z′-factor has been described as statistical parameter in HTS assays taking in account the variation between a signal in maximum response, the positive control, and minimum response, the negative control (27). Hence, the Z′ factor can be used as guideline to optimize semi-quantitative methods in order to improve discrimination between the positive and negative controls. Mean and standard deviation (SD) were determined for the positive control, inhibitor condition, and negative control, vehicle condition, to calculate the Z′ factor per experiment as follows:
$$ {Z}^{\prime }\ \mathrm{factor}=1-\frac{3\ \left(\upsigma\ \mathrm{positive}\ \mathrm{control}+\upsigma\ \mathrm{negative}\ \mathrm{control}\right)\ }{\left|\upmu\ \mathrm{positive}\ \mathrm{control}-\upmu\ \mathrm{negative}\ \mathrm{control}\right|} $$
Statistical Data Analysis
All statistics were performed using GraphPad Prism. Data is presented as mean ± SEM of at least three independent experiments (n = 3) and considered to be significantly different if p < 0.05 using an unpaired two-tailed Student’s t test, unless stated otherwise.