Results

A glomerular and proximal tubular microphysiological system

Filtration and reabsorption of the glomerulus and PCT were assessed by two main devices in the construction of the MPS (Figure 1C-D ). Within the stainless steel housing unit (Figure 1C ), the glomerulus enclosed a single porous PES membrane (porosity = 60) seeded with the differentiated CIHP-1 and HUVECs. The differentiated podocytes were cultured onto the GECM and EECM coated PES membranes 14 days prior to initiation of flow. The negatively charged GECM and EECM generates a charge-selective filtration barrier of the glomerulus (C. M. Sakolish et al., 2016). During differentiation, podocytes were observed for flattened morphological characteristics with finger-like projections in areas surrounding PES membranes. On day 15, PES membranes were flipped and seeded with the endothelial cells for 4 hours. Once the HK-2 cells were seeded onto a fibronectin-coated polycarbonate membrane (porosity = 20) for 4-hours, it was sealed in the machined bi-layer, polycarbonate microfluidic device described by previous work (Figure 1D ) (C. M. Sakolish & Mahler, 2017; C. M. Sakolish et al., 2019). Its physical dimensions (30 mm long x 15 mm wide x 50 μm thick) is comparatively similar to reported in vivo PCT dimensions in the human kidney (14 mm long x 40 μm thick) (Lote & Lote, 1994). The fully constructed MPS was connected to a T-Junction (Figure 2A-B ), a splitter of fluid flow to recycle 90% of flow into the bloodstream and 10% into the filtrate output. While the peristaltic pump maintained the MPS flow rate for the seven-day operation, additional modeling was required to optimize tubing length, diameter, and to safeguard physiological flow rates and shear rates (0.4 -1.5 dynes/cm2) within both glomerular housing unit and PCT microfluidic device.

Flow characterization of microphysiological system

To mediate the MPS microenvironment, a two-dimensional (2D) numerical simulation of porous media transport was performed using a finite element model of the device via COMSOL MultiPhysics®. A series of parametrizations (Table ) was performed for the velocity flow (m/s) (Figure 3A-B ) and shear rate (1/S) (Figure 3C-D ) based on the assumptions of steady state, constant fluid flow, viscosity and density, incompressible flow, minimal fluid evaporation, and negligible gravity. Conversions into physiological measurements were adopted using the following shear stress (Equation 1 ).
\(\tau=\frac{6Q\mu}{bh^{2}}\ \) (Equation 1 )
where μ is the endothelial serum free culture medium viscosity at 37 °C; \(\tau\) is the shear stress (dyn-s/cm2);Q is the volumetric flow rate (cm3/s); bis the channel width; and h is the channel height (Q = 7.6 x 10-4 cm3/s, μ = 8.9 × 10−3 dyn-s/cm2, h  = 0.05 mm,b  = 30 mm)
Based on the conversions, the optimal tubing length of filtrate output and bloodstream were ~508 mm and 734 mm respectively. The tubing length ranges were selected based on peristaltic pump and incubator sizing constraints. The inlet mass flow rate was 7.6 x 10-7 kg/s (7.6 x 10-4cm3/s) to generate a shear stress of 0.5 dyne-s/cm2 across the apical layer of the PCT device. Means of 1-D velocity profiles from filtrate output and the bloodstream had a ratio of 83/81 (Figure 3 ). Concentration profiles of 0.1 mg/mL FITC-HSA were computed using its diffusivity coefficient (61 μm2/s) in Transport by Diluted Species with a time-dependent solver from 0-120 minutes (Figure 3E-F ). The concentration 1-D profile was characterized with an exponential increase until it plateaued and approached steady state around 105-120 minutes. While these simulations are effective in determining the optimal geometry for the desired shear stress, the model was idealistic and reproduced without the inclusion of cellular effects that may alter the overall outcome.
Inlet mass flow rate (kg/s) was experimentally validated by using a single inlet 0.25 ID mm tubing at 4 -16 rpm pump rate (0.19 – 0.72 dyne-s/cm2) (Figure 4A-B ). Flow rates (μL/min) and shear stress (dyne-s/cm2) was determined with pumping 1X phosphate buffered saline (PBS) over the course of 2 hours, and collecting samples at 15-minute intervals (n= 4 for each rpm). At 12 rpm pump rate, the flow inlet (Qin) achieved an average of 45.3 ± 3.0 μL/min with a shear rate of 0.5 ± 0.04 dyne-s/cm2. With the assembled MPS, 1X PBS was cyclically pumped for the same duration, and filtrate formation (Qprimary filtrate) and bloodstream recycling (Qblood) rates were an average of 16.2± 4.5 μL/min and 29.0± 4.5 μL/min, respectively. Filtrate output and shear stress across the PCT for the single culture and tri-culture conditions (Figure 4C-D ) were compared with the simulated and experimental settings to characterize the system under non-ideal conditions. Single culture devices contained only the proximal tubule (HK-2) cells seeded into the PCT device, while the tri-culture was seeded with CIHP-1 and HUVECs in the glomerulus housing unit along with the HK-2 cells in the PCT. Implementation of single culture is representative of pre-established PCT devices (Homan et al., 2016; Jang et al., 2013; Jansen et al., 2016; Ng et al., 2012; Raghavan, Rbaibi, Pastor-Soler, Carattino, & Weisz, 2014; Sabbisetti et al., 2014; C. Sakolish et al., 2020; C. Sakolish et al., 2018; Weber et al., 2016), whereas tri-cultures reflect the combination of the human glomerulus and PCT. While the 90.0% of fluid was not recycled into the bloodstream as shown in the simulation, approximately 36.0% was outputted into the waste collection and 64.0% was recycled into the MPS.

Recapitulation of glomerulus and proximal tubule functions

Key characteristics of a functional nephron are its blood serum protein filtration, glucose resorption, and filtrate formation capacity. Two culture conditions were established to demonstrate system characteristic differences, which was denoted as a single (n=3) and tri-culture (all cell-types) (n=3). Throughout the seven-day operation, primary filtrate formation was collected every 24 hours in conical tubes and immediately stored at -20°C before additional assessment. Volumes of waste for each culture type were weighed to measure velocity flow rate (mL/min) (Figure 5A-B ). The primary filtrate formation in the single-culture maintained a median of approximately 30 μL/min, while the tri-culture had 15.0 μL/min. The introduction of the glomerular cells and endothelial cells created a barrier effect, reducing the flow rate within the apical PCT of the MPS to generate a shear stress from 0.3 to 0.2 dyne-s/cm2. Results of the simulated, experimental, single, and tri-culture filtrate output and shear stress across the apical region of the PCT were shown (Figure 4C-D ). While the simulated model exhibited a linear behavior based on its regression (R2=1.0), the shear stress under experimental conditions significantly decreased in the presence of cells, suggesting that simulations provided only ideal conditions and required further parametrizations.
Since the proximal tubule is responsible for nearly 90.0% of glucose reabsorption in the kidney, glucose secretion (mg/mL) in both single and tri-culture devices were evaluated using the Infinity Glucose Assay (Figure 5C-D ). Filtrate output volumes were sampled three times into a 96 well plate and compared to a 1 mg/mL glucose solution diluted with 0.1% benzoic acid standard curve. Samples were read with the plate reader at an absorbance of 340 nm. Both culture conditions lacked statistically significant difference due to the large volume of daily filtrate output (~30 - 40 mL/day), although there was a downward trend in the amount of glucose present in the filtrate output over the course of the 7 day culture, indicating some proximal tubule function. No significant change in both velocity and glucose concentrations (Figure 5B and D ) demonstrated that the MPS was stable during the seven days.
On Day 7 of MPS operation, 0.1 mg/mL of fluorescently tagged human serum albumin (FITC-HSA, Abcam #ab8030) diluted with 1X PBS was circulated into the MPS to assess glomerulus function and were analyzed for the amount of unfiltered serum proteins in both the filtrate outlet and blood stream circuit reservoir (McQuarrie, Shakerdi, Jardine, Fox, & Mackinnon, 2011). Samples were drawn from the filtrate outlet and blood stream circuit reservoirs every 15 minutes for 2 hrs and placed in a black, 96-well plate. Samples were analyzed for fluorescence with a Synergy 2 plate reader against a standard curve (Figure 5E-G ). Protein concentration in the bloodstream minimally declined from 0.1 mg/mL of FITC-HSA for both case conditions. In the tri-culture conditions, there was a statistically significant difference (p< 0.001) between filtrate output and bloodstream circuit output (Figure 5F ). FITC-HSA concentration in tri-culture in the blood circuit output (0.083 mg/mL) was 8 times greater than the filtrate output (0.011 mg/mL), suggesting that there is a functional filtration process in the tri-culture MPS. This behavior was not exhibited in the single culture conditions. Single-cell culture exhibited an increase in protein concentrations in the filtrate output, while the tri-culture devices had little protein reach the filtrate output (Figure 5G ). Absence of protein in the waste or filtrate stream suggests that the glomerular cells prevent the FITC-HSA from passing through the glomerular membrane, similar to the function of the glomerulus in vivo .
Fluorescent stains provided insight into the cellular phenotype following dynamic culture. The membranes, PCT (polycarbonate) and glomerular (PES) membranes from the MPS were removed, fixed with 4% paraformaldehyde (PFA), stained with phallodin 568) and Hoescht 33342, and imaged using a Zeiss confocal microscope at 40x magnification. F-actin cytoskeleton restructuring was imaged with phallodin 568, whereas DNA in the nuclei was counterstained with Hoescht 33342. To determine the cellular confluence, qualitative 2D orthogonal views and three-dimensional (3D) views (Figure 6A-F ) were obtained. Qualitatively, the HK-2 cells on the PCT membrane (Figure 6A-B ) demonstrated a monolayer behavior as previously shown (C. M. Sakolish & Mahler, 2017), and quantitatively represented by the 75.3% confluence and significantly greater cell count in comparison to podocytes and endothelial cells (Figure 6G-H ). The CIHP-1 cells on the glomerular membrane appeared flatter, fewer in cell count, and fewer F-actin across the plane due to a larger average cell area (831.1 μm2) compared to HK-2 (183.6 μm2) and HUVECs (353.7 μm2). From the 3D view ofFigure 6C , finger like projections also validated full differentiation (day 14). The endothelial cells were highly directional in the presence of flow, but attachment was limited by day 7 (Figure 6E-F ). This may be related to HUVECs poor ability to adapt to the dynamic crossflow culture conditions.