Creating a kidney organoid-vasculature interaction model using a novel organ-on-chip system

Kidney organoids derived from human induced pluripotent stem cells (iPSCs) have proven to be a valuable tool to study kidney development and disease. However, the lack of vascularization of these organoids often leads to insufficient oxygen and nutrient supply. Vascularization has previously been achieved by implantation into animal models, however, the vasculature arises largely from animal host tissue. Our aim is to transition from an in vivo implantation model towards an in vitro model that fulfils the advantages of vascularization whilst being fully human-cell derived. Our chip system supported culturing of kidney organoids, which presented nephron structures. We also showed that organoids cultured on chip showed increased maturation of endothelial populations based on a colocalization analysis of endothelial markers. Moreover, we observed migration and proliferation of human umbilical vein endothelial cells (HUVECs) cultured in the channels of the chip inside the organoid tissue, where these HUVECs interconnected with endogenous endothelial cells and formed structures presenting an open lumen resembling vessels. Our results establish for the first-time vascularization of kidney organoids in HUVEC co-culture conditions using a microfluidic organ-on-chip. Our model therefore provides a useful insight into kidney organoid vascularization in vitro and presents a tool for further studies of kidney development and drug testing, both for research purposes and pre-clinical applications.


Results
Microfluidic organ-on-chip system supports kidney organoid development. Kidney organoids were generated from iPSC of healthy donors over a 20 day differentiation period (Fig. 1A). On day 11 organoids were placed onto the culturing chamber of the organ-on-chip system and subjected to microfluidic flow by pumping of medium through the 3 microfluidic channels beneath. Organoids were harvested at day 20 (Fig. 1A). Organoids cultured on chip are subjected to submerged culturing in the chip system, and medium pumped through the three microfluidic channels reaches the culturing chamber through the porous membrane (Fig. 1B), contrary to transwell systems where the organoid is subjected to an air-liquid interface. Kidney organoids cultured on the chip system (Fig. 1B, C) showed expression of the glomerular marker WT1, podocyte marker PODXL, proximal tubuli marker Villin, and distal tubuli maker E-cadherin similar to organoids cultured on transwell membranes (Fig. 1D). These results confirm culturing of kidney organoids is supported by the microfluidic organ-on-chip system. Endothelial tissue in kidney organoids cultured on chip shows directed maturation patterns. To better assess the effects of the chip system on the vasculature, we performed a series of analy- www.nature.com/scientificreports/ ses focused on organoid vascular tissue maturation. We performed the analysis based on the area of MCAM + and PECAM + EC populations using immunofluorescent images of MCAM-PECAM co-staining ( Fig. 2A). Area analysis showed a two-fold increase of MCAM + and a four-fold increase of PECAM + areas in comparison to organoids cultured on transwell (Fig. 2B). Moreover, colocalization analysis of MCAM and PECAM showed organoids cultured on transwell had a higher percentage of ECs expressing non-colocalized MCAM and PECAM whilst the percentage of cells showing colocalization of both markers was low (Fig. 2C), (Supplemen-  HUVECs create 3D synthetic vessels on chip channels. We next aimed to induce migration of endothelial cells from the chip channels into the organoids. To do this, HUVECs were cultured in the microfluidic channels. HUVECs formed a confluent monolayer on all surfaces of the chip's channels after 48 h of static culture ( Fig. 3A) (Supplementary video 1, Supplementary video 2). HUVECs were still present in the form of a confluent monolayer 24 h after initiating flow (Fig. 3B, C). This confirmed the capacity of these cells to resist flow sheer stress. Flow was sustained at 3µL/min throughout the experiment, controlled by the flow sensor placed before the chips in the circuit. Upon withstanding flow for 24 h, HUVEC acquired directionality on the channels concurrent with flow direction as it is seen in vascular tissue architecture, suggesting the flow exercised over these microfluidic channels mimics native vascular conditions 19 (Fig. 3B, C, Supplementary Fig. 2). Moreover, 3D-rendering of the confocal stacks showed that HUVECs not only covered all planes of the channels creating a synthetic vessel, but cells also migrated through the porous membrane and formed a monolayer in the bottom plane of the culturing chamber ( Fig. 3D) (Supplementary video 1, Supplementary video 2). This effectively proved the capacity of HUVECs to migrate from the channels towards the culture chamber through the pores present in the membrane (Fig. 3D). To investigate HUVEC migration through the porous membrane we lined the chip channels with GFP + HUVECs and cultured kidney organoids above for 9 days (Fig. 4A). Upon collection of the organoids, organoid tissue contained GFP + HUVECs derived from the channels, forming vascular structures presenting an open lumen (Fig. 4B). Moreover, these GFP + structures connected with endogenous ECs (GFP -) from the organoid tissue as shown by PECAM immu- www.nature.com/scientificreports/ nostaining (Fig. 4B). We therefore demonstrated the formation of endothelial structures derived from a bioartificial vessel in kidney organoids in a microfluidic organ-on-chip.

Discussion
The present study demonstrates that kidney organoids can be successfully differentiated on an organ-on-chip and present all relevant nephron structures upon culture on chip, demonstrating the feasibility of this system for organoid culture and differentiation. Moreover, the implemented system facilitates organoid culturing protocols due to its automatized pump that enables culture without the need for medium changes every 48 h. Organoids cultured on chip showed higher expression of endothelial markers MCAM and PECAM in comparison with organoids cultured on transwell based on the area analysis. Furthermore, upon area colocalization analysis of these markers, we observed that kidney organoids cultured on chip show directed endothelial maturation. Organoids cultured on transwell showed higher expression of both non-colocalized markers, which would correlate with early and late stages of endothelial maturation, suggesting their endothelial populations do not display cohesive native tissue maturation patterns 10 . These results further confirm the fluidic flow supplied by fluidic devices is an important cue for EC maturation as shown in previous work 9,20 . Moreover, the implemented system presents certain advantages over previous devices shown to support kidney organoid culturing, by enabling perfusability of the channels and migration of HUVECs into the tissue that successively form vascular structures. These channels supported the formation of perfusable synthetic vessels upon HUVEC culturing. This established a proof-of-concept for the use of this organ-on-chip to mimic vasculature in vitro. The HUVECs resisted 9 days of fluidic flow shearing stress and adopt a directionality in concordance with native endothelium when subjected to unidirectional flow 19 , further proving the capacity of this endothelialized microchannels to mimic vasculature in vitro. Upon co-culture with the channels endothelialized with GFP HUVECs, organoid tissue presented GFP + HUVECs forming vascular structures that connected to native organoid ECs. These results demonstrate that co-culture of kidney organoids and HUVECs on the chip system leads to endothelial cell infiltration and vascularization under microfluidic flow. To our knowledge, we present the first successful infiltration of HUVECs inside kidney organoids where they contribute to vascular development. Our research constitutes a first step towards functional in vitro vascularization of kidney organoids. We hypothesize the high production of VEGF by the kidney organoids ( Supplementary Fig. 3) acts as a powerful cue for HUVECs' migration towards the tissue. This claim is supported by various publications reporting the role of VEGF to recruit ECs 8,21-23 , although additional factors may play an important role in HUVEC migration. These factors need to be further researched, as well as the mechanism that leads HUVECs to form vascular structures that are integrated with native organoid EC populations. Functional vascularization of organoids has previously been achieved in vivo by implantation in animal models, and has proved to aid in the maturation processed of kidney organoids 8,13,14 . These results confirm vascularization can improve kidney organoid maturation. However, reliance on animal models hinders the use of these models, as the vasculature derives largely from host tissue 14 , which hampers translatability of results in regards to human tissue. Ultimately, our goal is to transition from in vivo models of vascularisation to an in vitro model using an organ-on-chip based system, which allows us to generate a fully human-derived research model that can be perfused. In the future, we expect to further improve this model by optimizing the use of iPSCs-derived ECs to obtain a model for kidney organoid-vasculature interaction that presents the same genetic background and that could pose as a powerful tool for personalized medicine with applications such as patient-specific drug testing. The implemented system presents a series of limitations regarding organoid culture that need to be overcome through future optimization processes. Although we were able to obtain successful kidney organoids upon culture on chip, our protocol still relies on generating the organoids following traditional transwell protocols 7 . Therefore, the organoids are only subjected to flow for part of the differentiation period. Moreover, the system largely constrains the number of organoids that can be generated during one differentiation protocol, which slows down the data obtention process. Regarding the volume of medium necessary to culture each organoid, the use of a non-circulating medium set-up significantly increases the volume used, and therefore the cost/organoid. Finally, we must take into account a clear limitation of the kidney organoid field being reproducibility 4 . Although improved protocols to generate kidney organoids have emerged, reproducibility still poses as a challenging issue in the field 2,24 , meaning large variability of the results. A collaborative effort must be done in this regard amongst researchers in the field to ultimately achieve consistently reproducible kidney organoids.

Summary
Kidney organoids derived from human iPSCs are a powerful model to study kidney development and disease, however their lack of vascularization hampers their maturation and longevity. In this work we employed a novel organ-on-chip system to initiate vascularisation in kidney organoids. We proved that this organ-on-chip system supports kidney organoid differentiation and improves the number of endogenous endothelial cells as well as their maturation patterns. Moreover, we created synthetic vessels by seeding HUVECs in the microfluidic channels of the chip. When co-cultured with kidney organoids in the chip system, HUVECs from the channels were able to migrate through the porous membrane and establish vascular-like structures in the organoid tissue, which presented open lumens and connected to endogenous endothelial cells. This research demonstrates the first steps of in vitro vascularization of kidney organoids, which is of great value for a diverse range of applications such as developmental studies and drug testing. Organoid culturing. iPSC-derived kidney organoids were generated based on the protocol designed by Garreta et al. 7 (Fig. 1C). iPSCs were dissociated into single-cells via incubation for 5 min using TrypLE™ Select (Invitrogen, Netherlands) at 37 °C. Cells were plated as a monolayer on a 6 well-plate and E8 medium was supplemented with Rock inhibitor RevitaCell (Invitrogen Medium was then refreshed, and afterwards growth factors were removed by replacing the medium with advanced RPMI 1640 basal media without further additions. Medium was refreshed every other day for the next 9 days. On the last day of the protocol, organoids were collected and fixed in a 4% paraformaldehyde (PFA) solution, stained using a 1:100 tissue marking dye kit (Thermo Fisher Scientific) in DPBS solution and embedded in 1% agarose (Roche, Switzerland) before proceeding with immunohistochemistry and immunofluorescence.

Materials and methods
HUVEC culturing. Human umbilical cord endothelial cells expressing GFP constitutively were commercially obtained (product code ZHC-2402, Cellworks, San Francisco, USA). Cells were grown in EBM-2 endothelial cell growth basal medium (Lonza, Switzerland) containing 5% heat inactivated foetal bovine serum. Cells were passaged upon reaching 80-90% confluence using Trypsin-EDTA solution (Sigma-Aldrich) and incubation at 37 °C for 3 min. Cells were kept under passage 10 to ensure proliferative and angiogenic capabilities. The GFP expression of this commercial line was validated using flow cytometry and WT-HUVEC as negative control. Briefly, cells were detached using Trypsin-EDTA solution, resuspended in the appropriate volume of sheath fluid (Thermo Fisher Scientific) and analysed using the BD FACSCanto™ II Flow Cytometry Systems (BD biosciences, Belgium) ( Supplementary Fig. 4).
Organ-on-chip system. We acquired the BIOND organ-on.chip system (inCHIPit™ and comPLATE™) (BIOND Solutions B.V., Delft, The Netherlands). The inCHIPit™ is produced using polydimethylsiloxane film supported by a silicon frame, and consists of a culture chamber that communicates with the underlying three 400 µM-wide channels through 4 µM-wide pores 25 To generate flow, chips are placed on the comPLATE™ and this is connected to a pressure pump that generates 800 mbar of pressure, and pumps circulating medium from the reservoir through the tubing and towards the chip channels, producing an estimated flow rate across the porous membrane of 0.2 µL/min. The system allows to maintain a constant flow thanks to the pressure regulator placed in the system, which ensures continuous fresh medium and stable shear stress. These perfusable channels of the chip system allow us to maintain organoid culture subjected to fluidic flow, which is a key factor in itself for vasculature development 9,20 . The chips were subjected to plasma ashing for 2 min using the plasma cleaner PDC-32G-2 (Harrick Plasma, New York, USA) at high radio frequency setting equivalent to 18 W. After treatment, chips were sterilized with 70% EtOH and stored submerged in MilliQ water to avoid contact with air. Previous to cell seeding, chips were cleaned in 70% EtOH for 5 min followed by two cleaning steps of 5 min with DPBS. Thereafter, a 0.01% fibronectin solution was injected on the chip channels and incubated for 1 h at 37 °C. Channels were flushed twice with DPBS before seeding of the HUVEC to remove any excess fibronectin.
HUVEC seeding on chip channels. HUVECs were removed from their culture flasks by trypsinisation and centrifuged at 800 g for 5 min. The pellet was resuspended in EBM-2 medium at a concentration of 2 × 10 6 cells/mL. Empty 200 mL filter tips (Greiner Bio-One, NL) were placed at one end of every chip before cell seeding. Using the same tips, 100 µL of cell suspension was flushed through the channels using positive pressure. The upper chamber of the system was filled with 150 µL of EBM-2 medium. The tips were left on the chip system and the plate was incubated at 37 °C with 5% CO2 for 1 h to allow the HUVECs to attach to the channel surface. After one hour, tips were removed and replaced with new tips, one of which containing 200 µL of EBM-2 medium. The flow generated by passive medium flow towards the empty tip at the other end of the channel ensured the www.nature.com/scientificreports/ removal of unattached cells and cell debris. HUVECs were subjected to static culture for 48 h to ensure proper attachment and proliferation, replenishing the medium in the tips and the upper chamber after 24 h. After 48 h of static culture, the chips were connected to the flow setup. After static culture, the complate was attached to pressure-based flow control, Flow EZ™ (Fluigent, Germany). Microfluidic flow was applied at a constant rate of 3µL/min for 48 h before placement of the kidney organoids in the top chamber. Flow and pressure were monitored by using the AIO software in combination with a flow sensor (Fluigent, Germany) and a bubble trap was added to minimise bubbles. Pressure threshold was set at 200 mbar to avoid leakage. Imaging of the channels was performed using confocal imaging, tiled images and stacks. Stacks were rendered into a 3D figure using the image analysis software Fiji. To confirm directionality of the HUVECs after flow, we manually counted the number of cells in comparable portions of the channels before (48 h) and after flow (24 h flow, 72 h total). To assess directionality cells that showed an elongation and direction concurrent with the flow direction with a deviation not greater than 45° were considered as having acquired directionality.
On chip culturing of kidney organoids. Organoids were collected after 5 days culture in the transwell system and placed on the top well of the chip system. Two organoids were placed in each chip. Medium of the reservoir was changed to 30% EBM-2 + 70% advanced RPMI 1640 basal media to support co-culture conditions of kidney organoids and HUVECs. Flow rate was kept at 3 µL/min and organoids were collected after 9 days for fixation and processing. Immunohistochemical staining. Four-micron sections of formalin-fixed paraffin-embedded organoids were stained with haematoxylin and eosin according to manufactures instructions (Ventana Medical Systems Inc., Arizona, USA). Immunohistochemistry was performed with an automated, validated and accredited staining system (Ventana Benchmark ULTRA, Ventana Medical Systems) using ultraview or optiview universal DAB detection Kit. In brief, following deparaffinization and heat-induced antigen retrieval the tissue samples were incubated according to their optimized time with the antibody of interest (Supplementary table 2). Incubation was followed by haematoxylin II counter stain for 12 min and then a blue colouring reagent for 8 min according to the manufacturer's instructions (Ventana Medical Systems Inc., Arizona, USA). Healthy kidney was used as positive control tissues ( Supplementary Fig. 5).
Multiplex immunofluorescent staining. Co-staining of CD31 and CD146 with was performed by automated multiplex IF using the Ventana Benchmark Discovery (Ventana Medical Systems Inc., Arizona, USA). In brief, following deparaffinization and heat-induced antigen retrieval with CC1 (#950-500, Ventana) for 32 min the tissue samples were incubated firstly with anti-CD31 for 32 min at 37 °C followed by detection with Red610 (#760-245, Ventana). Antibody denature step was performed using CC2 (#950-123, Ventana) for 8 min at 100 °C. Secondly, anti-CD146 was incubated for 32 min at 37 °C followed by detection with FAM (#760-243, Ventana). Slides were washed in DPBS and covered with DAPI in vectashield. Antibody information and clonality can be found in Supplementary Table 2. All slides used in this project were randomly picked before imaging, and orientation of the organoids was undetermined after processing.
Computational image analysis. Quantitative Assessment of marker expression in kidney organoid sections was performed using the image processing software Fiji 26 . Expression area was calculated by delimitation of each individual organoid. Expression of each marker was calculated as a relative percentage of the total area. Colocalization analysis was performed with the Fiji plugin JACoP 26,27 for the markers MCAM and PECAM. Threshold was determined individually using Costes' automatic threshold 27,28 . Statistical processing of the results was performed using two-tailed unpaired t-test with Welch's correction with a confidence interval of 95% using GraphPad Prism version 8.0.0 for Windows (GraphPad Software, San Diego, California USA, www. graph pad. com).

Data availability
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.