Spheroids, organoids and kidneys-on-chips: how complex human cellular models have assisted in the study of kidney disease and renal ciliopathies

Kidney disease is one of the leading causes of morbidity worldwide, emphasizing the importance for physiologically accurate disease models. With most of the approved renal drugs failing to perform as well in human clinical trials as they did in animal testing, it is imperative that new and improved human-based models are developed to test these potential therapeutics. One option is to use patient derived cell lines, grown in both two-dimensional (2D) and three-dimensional (3D) structures, known as spheroids and organoids. Despite their contributions to the field, the lack of physiological accuracy, including the absence of fluid flow, and mechanistic effects in these 2D and 3D models means there is still room for improvement. Organ-on-a-chip (OOAC) technology offers itself as a potential candidate model to overcome these limitations. Over recent years OOAC technology has grown in popularity, with multiple organ systems, including lung, liver, and kidney described in the literature. In this review, traditional human cellular based models, including monolayer, spheroid and organoid models will be discussed. Human kidney-on-a-chip models will also be discussed, while exploring the advantages and potential limitations of this rapidly emerging field for the study of human kidney disease and drug testing.


Introduction
The kidneys are a pair of highly specialized organs containing an estimated 1 million nephrons, the functional units of the kidney (Bertram et al. 2011). Each individual nephron contains a bundle of capillaries known as the glomeruli housed in the Bowman's capsule where glomerular filtration occurs (Fig. 1). The primary filtrate passes out into the renal tubule, along the proximal convoluted tubule, loop of Henle and distal convoluted tubule and to the cortical and medullary collecting ducts (Fig. 1). Along the tubule reabsorption of water, electrolytes and nutrients occurs in a nephron segment specific manner. The final urine exits the kidneys from the collecting ducts into the renal pelvis, down the ureter, into the bladder and is voided via the urethra. Besides the production of urine, kidneys are involved in the maintenance of blood pressure via renin producing cells in the macula dense (Suzuki and Saruta 2004). The intricate architecture of the kidneys, including the highly heterogeneous cellular population ( Fig. 1) provides many opportunities for mechanistic regulation, and complications, on both a cellular and molecular level. High blood pressure and continuous hyperglycemia in diabetics and genetic predisposition are amongst the most common causes of chronic kidney disease (CKD) (Webster et al. 2017).
With 1.2 million people thought to have died from some form of CKD in 2017 (GBD Chronic Kidney Disease Collaboration 2020), it is imperative that the underlying disease mechanisms, as well as potential treatments are extensively researched. Historically, cellular experiments using commercially available immortalized renal cell lines, and animal models including mice (Ross et al. 2006;Kurbegovic and Ramsbottom et al. 2020), zebrafish (Huang et al. 2014), and primates (Tsukiyama et al. 2019), have been used to gain insights into the mechanisms of renal diseases including renal ciliopathies. This group of diseases are caused by dysfunctions, or the absence of primary cilia microtubule based cellular protrusions found on the surface of almost every mammalian cell (Satir and Christensen 2007) (Fig. 2). To date, there are now 35 primary renal ciliopathy diseases, with 187 established causative genes and 241 candidate genes for other primary ciliopathies (Reiter and Leroux 2017). Examples of primary ciliopathies which may include a renal phenotype include Joubert Syndrome (JBTS), Oral-facial-digital Syndrome (OFD), autosomal dominant or recessive polycystic kidney diseases (ADPKD and ARPKD respectively), nephronophthisis (NPHP), and Bardet-Biedl syndrome (BBS) (Fig. 3).
However, limitations including interspecies differences and translational difficulties during clinical trials (de Caestecker et al. 2015) have driven the shift to patient-specific studies. To achieve a human-specific approach, renal cells derived from both tissue biopsies (Murray et al. 2021) and urine (Molinari et al. 2018;Molinari et al. 2020;Srivastava et al. 2017a, b;Bondue et al. 2021) have been used to study kidney diseases. To progress knowledge further, patient derived cells have also been used to create more complex, self-organized three-dimensional (3D) structures.
In this review, we will compare, contrast, and discuss the current cellular models used to study human renal diseases. We will also explore the advances made in renal models, particularly the involvement of microfluidic chip technology and discuss the possible limitations (Fig. 4).

Isolation and selection of cell types for use in renal studies
Historically, the gold standard to study kidney morphogenesis, patterning and disease has been the implementation of animal studies, predominantly mouse and zebrafish. Recent advances in cell culturing methods have allowed for the substitution of animal models, for those generated using primary renal cell lines from multiple origins (Table 1). Over recent years, an increased emphasis has been placed on the ability to study human renal diseases without the need for animal models. One example of how this has been achieved has been the use of human pluripotent stem cells  Table 1) to generate kidney specific cell lines. These hPSCs can be isolated from multiple starting materials, including liver, kidney, and most frequently dermal fibroblasts. In 2019, a three-stage differentiation programmed was described, capable of producing five distinct terminal and two precursor renal cell lines from hPSCs (Hariharan et al. 2019). In brief, the procedure involved 4 days of metanephric induction, 4 days mesenchymal epithelial transition of nephron progenitors, concluding with 6 days of nephronal cell specification (Hariharan et al. 2019). Cell types generated from hPSCs using this method included mesangial, proximal tubular, distal tubular, collecting duct epithelial cells and podocyte precursors as early as 14 days (Hariharan et al. 2019).
Primary renal cells can be isolated from tissue biopsies. However, to obtain healthy kidney cells via biopsy is difficult, as this invasive procedure is not approved unless disease is suspected. Proximal tubule, podocyte and glomeruli cells have all successfully been isolated from nephrectomies (Ma et al. 2015;Sanchez-Romero et al. 2019;Mene and Stoppacciaro 2008). Broadly speaking, to isolate these cells, kidneys are first decapsulated by removing the renal cortex, before tissue samples are cut into 1 mm 3 pieces. Next, samples are gently washed, before undergoing enzymatic digestion via collagenase. The digested material is then sieved, and cell types selected for, before plating in the appropriate media. This protocol was used to generate the immortalized human kidney cell 2 (HK-2, Table 1) in 1984 (Detrisac et al. 1984), a cell line which has been used in numerous renal studies since.
Fibroblasts derived from skin biopsies have often been used to study ciliopathies as primary cilia can be induced following a short period of serum starvation (Pollara et al. 2022;Shamseldin et al. 2020). However, inconsistent proliferation, viability (Genova et al. 2019) and organ-specific differences are associated limitations when using fibroblasts to study renal ciliopathies. Surprisingly, spontaneously shed milk teeth from children have also proven to be a source of ciliated cell lines (Ajzenberg et al. 2015). However, viability, and accessibility are limiting factors for the widespread use of teeth as a source of ciliated cells, as tooth loss is age limited.
In response to this, urine has been used as an alternative source of renal cells for over 50 years (Sutherland and Bain 1972;Linder 1976;Felix et al. 1980). Another non-invasive source of primary renal cells is fresh urine samples, often referred to as liquid biopsies. Briefly, urine samples are centrifuged so that any cells present in the urine sample can be separated from the urine. From here, the cells are then washed, before resuspension for plating for expansion to study both renal diseases, and primary ciliopathies.
Urine derived cells have been used to study renal diseases including cancer (Elliott et al. 1976), and primary ciliopathies (Molinari et al. 2018(Molinari et al. , 2020Srivastava et al. 2017a, b;Ajzenberg et al. 2015;Ziegler et al. 2022). The use of human urine-derived renal epithelial cells (hURECs) in 2D culture has led to the discovery of disease specific ciliary phenotypes including the elongation of primary cilia in JBTS hURECs with CEP290 mutations (Srivastava et al. 2017a, b); elongated and swollen primary cilia in ARPKD due to PKHD1 mutations (Molinari et al. 2020); and an accumulation of IFT88 at the tip of primary cilia in hURECs derived from a patient with novel missense variants in IFT140 (Oud et al. 2018). Rather surprisingly, these urine derived cells have also been used to study aspects of cardiac (Steinle 2019), vascular (Liu et al. 2018), liver (Hu et al. 2020), and neurological diseases (Guan et al. 2014;Stao et al. 2019).
Selection of cell-line origin can also prove influential on model success. Primary cell lines typically tend to have limited self-renewal potential, with increased variation found between donor types, resulting in difficulty with reproducibility. However, primary cell lines can be extremely valuable when considering disease drivers on a patient-specific basis. In fact, the use of primary cell lines has allowed for the identification of novel mutations, and the characterization of disease-specific phenotypes in primary ciliopathy patients (Molinari et al. 2018;Devane et al. 2022). In comparison, immortalized cell lines can offer numerous advantages to primary cell lines including: unlimited supply (as they are commercially available), easy to use, contain a pure population of cells, are consistent and generate reproducible results, are cost effective, and tend to be clear of infectious or hazardous pathogens (Kaur and Dufour 2012).

Two-dimensional (2D) cell models to study kidney diseases
The 2D culture of human cell lines provides data that is easily translated, uses simple methods, is low in cost, and has the potential to be upscaled to high throughput with the possibility to study multiple diseases, mechanisms, and treatments in short time frames. Furthermore, most cellular based assays available on the market have been developed specifically for 2D based models, reducing time spent optimizing protocols. Numerous advances in diseases with Fig. 3 Wheel depicting primary ciliopathies which present with a renal phenotype. Other common phenotypes associated with each primary ciliopathy are also shown. Created with BioRender. com renal phenotypes such as JBTS (Latour et al. 2020), ADPKD and APRKD (Wilson et al. 1986), and nephronophthisis (NPHP) (Ziegler et al. 2022) have been made using primary cells grown under 2D conditions. Relating to the study of ciliopathies, both dermal derived fibroblasts, and hURECs (Table 1) have been utilized in the generation of numerous 2D models. However, despite the major contribution to the field of renal diseases, these urine-derived 2D models face limitations. Examples of such include the fact that cellular differentiation is heavily reliant upon not only cell-cell interactions, but also a cells interaction with its microenvironment, known as the extra-cellular matrix (ECM) (Majo et al. 2020). The renal ECM is critical for phenotypic features, cell growth, differentiation, and gene expression (Makino et al. 2000;Bondue et al. 2021); this key component is often absent in traditional 2D-based models.
To addresses this issue, Sendai Virus reprogrammed hPSCs (Table 1), were grown on various extracellular matrix (ECM) coatings including ECM harvested from decellularized human foreskin fibroblasts, ECMatrix ™ -511, recombinant human laminin-521 and truncated recombinant human vitronectin (VTN-N), all under 2D conditions (Murphy et al. 2022). Following growth on these coatings, pluripotent markers were found to alter between each. Those hPSCs grown on foreskin fibroblast coatings were found to have increased levels of the dermal fibroblast marker HSP47, while those cultured on commercially available ECMs showed increased levels of pluripotent markers Hoechst, Nanog and POU5F1 (Murphy et al. 2022). Following the initial ECM characterization tests, hPSCs were then differentiated into podocyte-like cells and stained with specific markers (synatopodin and WT1) and following principial component analysis (PCA) showed good clustering of all samples except undifferentiated hPSCs (Murphy et al. 2022). Gene expression analysis also supported this, with high levels of podocyte-specific markers PALLD and COL4A3 (Murphy et al. 2022). This study highlights the benefits of using 2D models to study the human kidney; by adding an ECM coating to assist in cell differentiation, it is possible to generate desired renal cell types. Although the main aim of this study was to identify non-animal alternative sources of ECM coatings for hPSC differentiation into podocytes, the methods described could be further adapted and optimized for a range of applications including understanding kidney disease drivers, and potentially investigating drug toxicity in podocytes in 2D cultures.
The effects of the addition of fluid flow stress (FFSS) and tensile stress have been reported in 2D podocyte cultures to study congenital anomalies of the kidney and urinary tract (CAKUT) (Srivastava et al. 2017a, b). Hyperfiltrationmediated injury is one of the main causes of nephron loss in Fig. 4 Overview of the current cell-based models used to study renal disease. Cells can be derived from individual patients' via either a tissue biopsy, or a liquid biopsy (urine samples). Once cells are established in traditional monolayer cultures, they can then be seeded on/into Matrigel scaffolds, or specialised chips to generate 3D cell models. Created with BioRender.com those with CAKUT, resulting in glomerular injury. To study this, it is vital to replicate stress conditions on podocytes. Tensile (or compressive) stress was achieved by culturing podocytes on flexible membranes of specially constructed culture dishes (to mimic capillary walls), where negative pressure (via vacuum) was applied to stretch podocytes (Srivastava et al. 2017a, b). It was found that tensile stress was responsible for the formation of actin-rich centers and radial stress fibers, upregulation of COX-2 and EP 4 , and activates p38-MAPK and ERK1/2 pathways (Srivastava et al. 2017a, b). In comparison FFSS is applied directly to the surface of cells, in this instance specifically to the slit diaphragms of podocytes. In this study FFSS was applied following the development of a flow chamber where a flow rate of 0.2 dynes/cm 2 was applied (Srivastava et al. 2017a, b). FFSS low rates ranging from 0.015 and 1.75 dynes/cm 2 with high rates ranging between 8 and 649 dynes/cm 2 have previously been reported with podocytes in culture (Huang et al. 2012;Dandapani et al. 2007;Friedrich et al. 2006). Changes induced by FFSS included disruption of actin stress fibers with the formation of a cortical actin ring, upregulation of COX-2 and EP 2 , increased PGE 2 levels, and the activation of Akt-GSk3β-β-catenin and c-src/PLD1/mTOR pathways (Srivastava et al. 2017a, b).
Despite the addition of ECM coatings, and application of stresses to generate 2D renal models, there remains limitation when considering other aspects of renal studies. For example, in relation to drug toxicity related studies, it has previously been shown that drugs shown to have high efficacy in 2D in vitro models, when translated into patients have low efficacy (Shoemaker 2006). This is a limitation associated with the lack of complex 3D structure, diffusion gradients, and non-representative drug metabolisms (Antoni et al. 2015).

Three-dimensional (3D) cell models to study kidney diseases
To resolve these 2D model-based issues, research has turned to the use of 3D models including spheroids and organoids. Throughout the literature, the term spheroid and organoid has been used interchangeably, often resulting in some confusion when discussing models (Simian and Bissell 2017). In general, spheroids hail from single-cell lineages and are described as free-floating self-organized spherical cellular aggregates, with a low level of complexity (Lancaster and Knoblich 2014). In comparison, organoids are considered more complex than spheroids, not only in their physical structure but also their definition, with numerous unique cell culture techniques described as organoids in the current literature (Simian and Bissell 2017). Organoids tend to as form structural units which are more accurate in resembling a specific organ structure as well as function (Lancaster and Knoblich 2014). Spheroids and organoids have both their own unique and some overlap in their purposes, protocols, and origins. It is for these reasons, that caution is advised when using the terms spheroid and organoid to prevent confusion. However, due to the benefits of 3D models, including the replication of cellular organization, signaling networks, differentiation and drug screening success, spheroids and organoids are growing increasing popularity in the renal field.

Renal spheroid models for studying disease
In 2011, human kidney epithelial cells (hKEpCs, Table 1) were described to aggregate into 3D spheroid structures under non-adherent conditions (Buzhor et al. 2011). These hKEpCs could maintain a proportion of renal developmental and stem-cell like markers when compared to those cells grown in 2D (Buzhor et al. 2011). Furthermore, renal spheroids have successfully been established from dissociated and undifferentiated hPSCs seeded between layers of Matrigel (Freedman et al. 2015). These hPSC derived renal spheroids expressed octamer-binding transcription factor 4 (OCT4), sex-determining region Y box-2 (SOX2) NANOG and TRA-1-60 pluripotency markers (Freedman et al. 2015).
Once differentiated into a tubular identity, spheroids were found to express the renal specific markers lotus tetragonolobus lectin (LTL), Lin11-Is11-Mec3 (LIM) homeobox 1 (LHX1), paired box gene 2 (PAX2), and E-cadherin (Freedman et al. 2015). Together, the described findings demonstrate the importance for consideration of model selection when wishing to investigate specific renal markers for disease purposes. RNA sequencing (RNA-seq) of 3D nephrospheres cultured on plates pre-coated with poly(2-hydroxyethylmethacrylate) ECM, derived from 2D hKEpCs, found significant gene ontology (GO) enrichment in gene sets involved in the differentiation of the epithelial segments of adult human kidney (Harari-Steinberg et al. 2020). Nephrospheres were also found to have enrichment for various nephron specific markers, and differentiation towards a renal epithelial phenotype after 1 week in culture (Harari-Steinberg et al. 2020). From a clinical perspective, following implantation of human nephrospheres into mice with CKD, cell engraftment, cell survival, proliferation, organization, and differentiation processes were all observed, implying a tubulogenic and therapeutic potential for nephrospheres (Harari-Steinberg et al. 2020).
For a patient-specific model, human urine-derived renal epithelial cells (hURECs) can be used to generate spheroids with apicobasal polarity, cilia formation, and complete formation of lumen (Ajzenberg et al. 2015). Briefly, hURECs were isolated via centrifugation of urine samples, with established cells then seeded into Matrigel. After 3-5 days culture in Matrigel fully formed spheroids were visible. The study also found that hURECs derived from healthy controls could form cilia in both 2D and 3D conditions, whilst hURECs derived from JBTS patients could only form cilia in 2D conditions (Ajzenberg et al. 2015). The significance of this finding highlights the advantage of 3D cell systems for the study of renal ciliopathies via the use of urine derived cells. From a disease specific perspective, hURECs have also been used to generate spheroids from CEP290 JBTS cells, with a partial rescue of ciliary defects following Hedgehog agonist treatment (Hynes et al. 2014).
More recently, a unique method of renal spheroid formation involving the incorporation of immortalized renal proximal tubule epithelial cells (RPTEC), and fibrinogen as a bioink was reported (Tröndle et al. 2021). This bioink was deposited onto a hydrogel layer of Matrigel, Collagen I and Fibrin before sealing with a second layer of hydrogel (Tröndle et al. 2021). This method allowed for the control of epithelial cell self-assembly into spheroids, with individual droplets leading to the formation of individual cellular clusters removing the need for complex 3D handling techniques (Tröndle et al. 2021). Depending upon the concentration of bioink used, the group observed the formation of tubules as early as day 1, with lumen typically forming around day 4 (Tröndle et al. 2021). Furthermore, following RNAsequencing, 31 out of 34 kidney-specific genes were found to be up-regulated in those spheroids which were bioprinted, compared to those cells grown in 2D conditions (Tröndle et al. 2021). Of those up-regulated genes, GO enrichment analysis showed enrichment in fibroblast growth factor and organic anion transport, two terms which are thought to be associated with tubular kidney structure formation in vivo (Tröndle et al. 2021). Collectively, these findings highlight the benefit of using complex cellular models for the study of renal conditions, as they can retain renal specific signatures which are often lost in simpler 2D models.

Renal organoid models to study disease
Early renal organoid models are present in the literature from around 2015 (Takasato et al. 2015) and are becoming more common. These 3D cellular structures have been used in the study of NPHP (Forbes et al. 2018), Wilms Tumors (Frans Schutgens et al. 2019) and ADPKD (Cruz et al. 2017;Kuraoka et al. 2020). Renal organoids, paired with singlecell RNA sequencing, have allowed for the characterization of key disease specific interactions indicating future genetic disease-specific research focus (Low et al. 2019;Wu et al. 2018).
Kidney micro-organoids containing around six to ten nephrons, surrounded by endothelial and stromal populations have been described in suspension cultures (Kumar et al. 2019). This study found that following extended time in culture (from 7 days onwards), cyst formation, and mesenchymal expansion was present with variation dependent upon cell line of origin (Kumar et al. 2019). This finding was consistent with other suspension grown micro-organoids (Cruz et al. 2017;Czernicki et al. 2018;Przepiorski et al. 2018). The use of these structures can assist in furthering the understanding of the driving factors of renal cyst formation, and potentially be used to identify therapeutics for the treatment and potential prevention of cysts, without the need for animal models.
Cyst formation within organoid structures has been used to model polycystic kidney diseases (PKD), including ADPKD which has previously been attributed to homozygous mutations in PKD1 (Cruz et al. 2017;Kuraoka et al. 2020). Upon the introduction of truncating PKD1 and PKD2 genes into kidney organoids derived from hPSCs spontaneous cyst formation was observed (Freedman et al. 2015). The presence of fluid accumulation, and proliferative epithelial cells lining the cystic region accurately replicated the ADPKD cellular phenotype in organoid models (Cruz et al. 2017). Following forskolin treatment of organoids derived from the iPSCs of an ADPKD and gene edited PKD1 mutant cells, cyst formation was observed and used to model the early stages of cystogenesis in ADPKD (Kuraoka et al. 2020). In comparison, ARPKD is thought to arise because of mutations in the PKHD1 gene (Sharp et al. 2005). ARPKD renal organoids generated from patient iPSCs have also been used to study the dilation of cysts (Low et al. 2019) and the ureteric epithelium cyst formation of ARPKD (Howden et al. 2021). Treatment of both ADPKD and ARPKD patient-derived renal organoids with the cystic fibrosis transmembrane conductance regulator (CFTR) was shown to block cyst formation (Low et al. 2019;Shimizu et al. 2020). Again, these studies show potential for the use of renal organoids as disease screening tools for the study of cyst formation.
Podocytopathies have also benefited from the use of patient-derived cells to form organoids, for the study of congenital nephrotic syndrome (Romero-Guevara et al. 2020). Following CRISPR/cas9 gene editing in hiPSC-derived organoids was found to restore podcyte transcriptional profiles whilst also rescuing NPHS1-associated disease phenotypes (Hale et al. 2018). Podocyte developmental pathways have also been investigated using human derived organoids. For example, podocalyxin was found to be crucial to providing cells with a negative surface charge (Freedman et al. 2015;Kim et al. 2017). Epithelial lumen formation, and podocyte tight junction formation, as well as microvilli formation, are all heavily reliant on the presence of negative charges on podocyte cell surfaces (Freedman et al. 2015;Kim et al. 2017).
Another example of a genetic kidney disease studied via organoid models include Mucin 1 kidney disease (MKD) (Dvela-Levitt et al. 2019). Caused by a frameshift in MUC1 (MUC1-fs), MKD is characterized by the inheritance of tubulo-interstitial kidney disease, with patients requiring either dialysis or kidney transplantation during their lifetime (Kirby et al. 2013). After treatment via BRD4780, it was proven that mutant MUC1-fs protein levels were cleared from intracellular compartments of iPSC-derived patient organoids (Dvela-Levitt et al. 2019).
Wild type iPSCs have previously been differentiated into mesoderm cells under 2D conditions (Morais et al. 2022). These mesoderm cells were then differentiated into kidney organoids, positive for glomerular and tubular structures after 18 days in organoid culture (Morais et al. 2022). Following immunofluorescence, basement membrane sequence of formation was confirmed in these organoids as suitable for investigations into renal BM developmental pathways (Morais et al. 2022).

Renal organoid models for drug screening
Renal organoids have also been utilized in the study of drug-induced nephrotoxicity. Assessment of renal-specific injury markers within kidney organoids can help indicate the response of cells to drugs. Kidney injury marker 1 (KIM1), and tubular epithelium cell death, has previously been induced following gentamycin and cisplatin treatments in hPSC-derived kidney organoids (Morizane et al. 2015). The loss and damage of podocyte cells within has also been found to be caused by the treatment of hPSC-derived organoids with Adriamycin (Kumar et al. 2019;Lawlor et al. 2021). Organoid-derived glomeruli (OrgGloms) and primary podocytes cultured from OrgGloms (referred to as OrgPods) offer increased renal toxicity screening capacity, with emphasis on podocyte toxicity (Hale et al. 2018). Collectively, these studies highlight how 3D cellular models are greatly beneficial in the understanding of drug induced kidney injury.
One of the most common limitations associated with 3D culture is the formation of necrotic centers, due a lack of oxygen when structures increase in size (Grebenyuk and Ranga 2019). Furthermore, efficiency, lifespan, and repeatability of 3D structures is also more variable when compared to 2D studies (Hickman et al. 2014). Another important consideration to make when using ECMs such as Matrigel and collagen-based products is the presence of potential behavior altering components in the scaffolds. For example, Matrigel, which is derived from Engelbreth-Holm-Swarm tumors in mice, was previously found to contain growth factors including fibroblast, and epidermal growth factors (Vukicevic et al. 1992). Furthermore, following global transcriptomic analysis of those organoids derived from hPSCs, it was concluded that these organoids are accurate representations of first and second trimester fetal kidneys, which may pose limitations when considering renal diseases occurring during a more mature state (Takasato et al. 2015).
Recently, the combination of spheroids, organoids and microfluidic chips have enabled great contributions to the renal field, furthering our understanding of key mechanisms including development, maturation, and responses to drugs. The addition of high flow shear stress (FSS) for 10 days perfusion, to kidney organoids grown on 3D-printed microfluidic chips was found to assist the formation of vascular networks, and nephron formation (Homan et al. 2019).

Organ on a chip
The basic component of modern organ chip technology involves the presence of channels, which can allow seeded cells to be tested under microfluidic conditions (Kimura et al. 2018). Despite the numerous chip structures, cell lines, and flow rates used in chip technology, there remains five key common possibilities shared by them all. These are: (1) replication of tissue-specific architectural arrangement, (2) culturing of multiple cell types to capture cellular interactions, (3) incorporation of mechanical cues, (4) ability of cell sensing, and (5) drug stimulation or delivery (Valverde et al. 2022).
Perhaps one of the most attractive qualities of chip technology for the study of renal disease is point 3, in the form of replicating urinary flow and influencing cellular morphology using flow mediated shear stress (FSS) and mechanical strain (Maass et al. 2019). Furthermore, the presence of FSS is essential for the trafficking and expression of apical and basolateral transporter-proteins in renal drug pathways (Petrosyan et al. 2019).

Chip material
There are multiple types of microfluidic chips commercially available, whilst some research groups prefer to design and fabricate their own depending upon the experiments they wish to conduct. One of the first things to consider when selecting a microfluidic chip is the material it is made from, and the associated advantages and limitations of said material ( Table 2). Choice of chip material is dependent upon multiple factors, the functionality of device, biocompatibility, read-outs, and fabrication strategies. Some of the most used materials used include poly(dimethylsiloxane) (PDMS), glass and thermoplastics (Leung et al. 2022).
PDMS is the most used of all these materials, mainly due to its high biocompatibility, gas permeability and optical transparency. Mechanical stimuli can also be replicated using PDMS chips, due to their relative flexibility, making it a good candidate for use in kidney-on-a-chip models (Leung et al. 2022). However, one of the main limitations of PDMS chips is their permeability to small hydrophobic molecules (Grant et al. 2021), which can make it difficult to predict drug responses, potentially hindering the full effectiveness of these chips for renal drug toxicity studies. Efforts have been made to try and reduce this small molecule absorption, including coating PDMS channel surfaces with titanium dioxide ( Another commonly used chip material is glass. Unlike PDMS chips glass does not absorb small molecules making it an ideal choice for use in drug screening models. In 2019 double chambered glass chips were fabricated by layering two glass plates with wet etching, sandblasting, computer numerical control (CNC) machining and thermal bonding. Briefly, wet etching and sandblasting were responsible for producing channels 0.4 mm deep, while CNC was responsible for the engraving of cell culture wells onto one glass plate (Hirama et al. 2019). Shallow channels to be used as Laplace vales 27 μm deep were generated using via wet etching, with plates combined to form the chip using thermal bonding (Hirama et al. 2019). The same study investigated the variability of flow rates through glass and PDMS chips; a constant flow of 4 kPa was applied to chips' chambers, with glass chips found to have a higher flow rate (Hirama et al. 2019). This variability on flow rate was attributed to the flexibility of chips, further highlighting the importance of material consideration when designing renal-based models, as flow rates can influence cellular differentiation.

Chip treatments and coatings
Chip surfaces may require certain treatments to enhance biocompatibility with cells, or even cellular adhesion. Those chips intended for use with spheroids and organoids often undergo pluronic acid treatments to prevent dissociation caused by cell adherence to chip surfaces (Leung et al. 2022). In other chips, where cell adherence is desired, surfaces may be coated with ECMs, such as Matrigel, fibrin, or collagens (Table 3). The most used ECM components for kidney chip models contain a collagen (either Type I or Type IV), gelatin or fibrin (Table 3). It is also important to Mimicking of the podocyte-GBMendothelial interface, and the urinary and capillary compartments of the glomeruli under flow conditions (replicated using cyclic suction).
Cells differentiated under flow conditions were found to express greater levels of nephrin against static cells suggesting mechanical flow contributes to podocyte differentiation (Musah et al. 2017)  had their channels washed with ethanol before being coated with rat tail-derived collagen I supplemented with DPBS, HEPES and genipin and NaOH to give a final pH or 8.0-8.5. Following an overnight incubation at 37 °C, both aligning fiber pins were removed to form tubular collagen lumen. Circuits were perfused with culture medium, using the Nortis pump set to 1 μL/min. The following day, one channel was injected with collagen IV in DPBS to coat the lumen before incubation at 37 °C. After cells were seeded into the respective channels, media was continuously perfused through each channel at a rate of 1 μL/min (FFSS of 0.9 dyne/cm 2 ) for 7 days Cells cultured after 7 days in the Nortis chips were positive for the tight junction protein ZO1, and primary cilia localised to the apical membrane showing a brush border. Increased expression of SLC47A1, SLC47A2 was also observed (Nieskens et al. 2020) Kidney on a Chip Human kidney tissue-derived cells, epithelial origin Nortis device coated with an ECM consisting of collagen I, while cell chambers were coated with collagen IV to facilitated cell adhesion. Channels containing cells were perfused at a rate of 0.5 μL/min Cells grown in chips were found to have EC 50 values fourfold higher than cells grown in monolayer. Expression of KIM-1 was also found to be fourfold higher in cells grown on chips. Kidney injury biomarkers calbindin, clusterin, oseoactivin, KIM-1, VEGF and α-GST were also significantly different. Collectively these findings suggest that chip models are more accurate for drug screening (Weber et al. 2018) consider the cell line, and disease, which is to be modelled in the chip, as this will undoubtably influence the ECM of choice (Fig. 5).
The kidney consists of three main ECM regions; the glomerular, tubulointestinal and vascular (Fig. 5). In general, the glomerular basement membrane is thickest, containing four macromolecules laminin, collagen IV, nidogen and heparan sulphate proteoglycans (Bulow and Boor 2019;Genovese et al. 2014). The glomerular basement membrane is exposed to ultra-filtration and would explain why those models wishing to study glomerular diseases would require a stiffer and thicker ECM.
Other matrix and scaffolds used for chip coating have included alginate with myeloblasts in drug metabolism and anticancer studies (Sung and Shuler 2009), Poly(D,L-lactide-co-glycolide) with oral squamous cell carcinoma (Fischbach et al. 2007), and foamed polylactic acid used with hepatocytes and brain cancer cells for anticancer activity studies (Ma et al. 2012).

Mechanical stimuli in chips
In the human body, kidney cells are exposed to multiple dynamic biomechanical stimuli including the increase/ decrease of ECM or basement membrane stiffness, changes in topography, geometrical confinement, and stresses such as fluid flow stress (FFSS), tensile stretch, and compression (Wang et al. 2022). Mechanical stimuli in the human kidney can be classified as either passive or active depending upon the characteristics (Fig. 6).

Passive stimuli
Passive mechanical stimuli are provided to cells in the kidney from their ECM network, including both signaling pathways and structural support allowing for cellular regulation. Other contributing factors of passive mechanical stimuli include the substrate stiffness, the surface topography, composition, and confined geometry of the ECM. In health, the glomerular ECM elasticity is typically 2.5 pKa, however in cases of disease, the stiffness decreases (Wang et al. 2022), highlighting the importance of the correct ECM stiffness for kidney-disease chip models. In relation to surface topography, both proximal tubule and podocyte cells tend to adhere to surfaces with a slight curvature, as apparent in their native curved basement membrane. Curvature of substrates has been found to be influential in renal cell behaviors (Yu et al. 2018), again highlighting the importance of this considering when designing renal-chip models. Lastly, the confined geometry of the cellular environment also influences renal cell behaviors. Typically, adherent cells spread

Active stimuli
Active mechanical stimuli include fluid shear stress (FSS), compressive pressure, and cyclic stretching. Perfusion of media is one of the main hallmarks of OOAC technologies, offering biological advantages to studies by mimicking urinary flow in kidney models, and blood circulating in heart models. Cell lines including glomerular endothelial, proximal tubule, distal tubule and podocytes have all been shown to be sensitive to FSS. In chip-based models, FSS has been applied to single channels, as well as more complex chamber-based chips containing apical and basal chambers for improved renal architecture replication (Wang et al. 2022). When designing a renal-chip model, it is also important to establish the flow rate to be applied, for example those kidney diseases involving hypertension typically have a higher flow rate applied. Another key active stimulus that is important to consider is compressive pressure and stretch. In the glomerular filter, a drop in pressure creates compression and stretching in the capillary walls, while hypertension has been shown to create podocyte loss due to increased mechanical strains (Endlich et al. 2001).

Chips to study kidney disease
In their most basic form, renal organ-chips involved the adherence of MDCK and human kidney-2 (HK-2, Table 1) cells to the base of microchannels under physiological shear stress (Huang et al. 2013;Zhou et al. 2014;Frohlich et al. 2012). These first initial humanized kidney on chip models showed that whilst under shear stress, cells form cilia, increased cell thickness and Na/K ATPase activity. Since then, more advanced renal organ chip model designs have incorporated double layers, with cells divided by porous membranes to study physiological responses to alterations in sodium concentrations and osmotic pressure (Jang and Suh 2010;Jang et al. 2013). Again, the formation of cilia, cell orientation, P-glycoprotein expression, cell polarity and albumin/glucose absorption were all influenced by the addition of shear stress (Jang and Suh 2010;Jang et al. 2013). Kidneys on chips models have also been used to study toxicity (Jang et al. 2013;Li et al. 2017;Chang et al. 2017), diabetic nephropathy (Wang et al. 2017), hypertensive nephropathy (Zhou et al. 2016), and virus-related renal dysfunctions (Wang et al. 2019). These studies offer exciting prospects for studying renal diseases, including ciliary defects in cells derived from ciliopathy patients on chip models. Renal chip technology has grown with rapid pace, with examples summarized in Table 2. However, further refinement and optimization may still be needed before the use of patient samples is possible, due to variation in success of cell lines cultured on chips.

Chips to study glomerular disease
The human kidney consists of multiple nephron segments (Fig. 1), each with its own unique role. Kidney on a chip technology has allowed for the study of each of these segments. The glomeruli are vital for the ultrafiltration of blood passing through the kidneys. Chip technology has been utilized to replicate the glomeruli filtration barrier (GBF) with podocyte and epithelial cell lines (Petrosyan et al. 2019;Iampietro et al. 2020). The glomerulus on a chip model have also been utilized to study hypertensive glomerulopathy (Zhou et al. 2016), early-stage diabetic nephropathy (Wang et al. 2017) and drug responses (Petrosyan et al. 2019).
Leading to end stage renal failure, hypersensitive nephrology is a common kidney disease with glomerular hypertension a driving force for disease progression. A triple layer chip, containing a porous polycarbonate membrane coated with basement membrane extract (BME) including laminin, Fig. 6 Overview of passive and active stimuli which can be applied and examples of how they can be used in models. Passive stimuli include the ECM characteristics of substrate stiffness, surface topography, ECM composition and the confined geometry. The second pas-sive stimulus chip model included is diffusion through membranes, as this is a passive process. Active stimuli include the compression and stretch of chips, as well as fluid shear stress. Created with BioRender. com collagen IV, entactin and heparin sulphate proteoglycan was designed to mimic the podocyte ECM (Zhou et al. 2016). Top channels of the chip represented the glomerular capillaries, while the bottom channel represented the Bowman's capsule (Zhou et al. 2016). Hypersensitive nephropathy was mimicked by infusion culture medium at a flow rate of 10 μL and 15 μL/min through the endothelium channel of the glomerular chip (Zhou et al. 2016). Following infusion, increased numbers of stress fibers were observed, with cell structure and shape also found to become irregular (Zhou et al. 2016). This glomerulus-on-a-chip proved capable of mimicking the glomerular microenvironment, with good biomarker expression, representative of the human kidney.
Glomerulus-on-a-chip (GOAC) wherein human podocyte cells and glomerular endothelial cell chip cocultures have been utilized to investigate injury responses to puromycin (Petrosyan et al. 2019) (Table 3). In this study, the GOAC was successful in maintaining podocyte/glomerular specific phenotypes, as well as being used to model kidney injury following exposure to puromycin aminonucleoside (PAN) (Petrosyan et al. 2019). One hour after exposure to PAN, cytoskeleton rearrangement and permselectivity for albumin of podocytes was observed (Petrosyan et al. 2019) highlighting how GOAC can be utilized for kidney function and nephrotoxic drug induced injury models.
Collectively, these glomerulus-on-a-chip studies offer the chance to advance management and treatment of glomerular diseases, including those patients with the poorest of prognosis. Two of the main constraints faced when studying glomerular diseases, including glomerulus-on-a-chip, is the maturation of the filtration barrier and the actual culture of podocyte cells themselves.

Chips to study tubular disease
The most abundant example of kidney-on-a-chip models are those which are derived from proximal tubule cells. Dualchannel microphysiologic chips seeded with HRPTECs were shown to express epithelial expression profiles for the tight junction cell-cell interactions, along with cilia formation into the tubule lumen (Nieskens et al. 2020). Human proximal tubule epithelial cells (RPTECs) have been successfully cultured under flow conditions within a luminal space of fibrinogen and thrombin coated chips (Ng et al. 2013). These cells were found to express proximal tubule markers as well as ion transporters; cell polarization was also observed (Ng et al. 2013). Following insulin recovery studies, leak proof barriers were found in those chips seeded with cells compared to those without (Ng et al. 2013). The addition of an Na + K + ATPase inhibitor was found to reduce the rates of reabsorption in cell laden chips (Ng et al. 2013); this finding could help understand how absorption of key molecules and therapeutics could behave within the kidneys.
More recently, advancements in both tissue engineering and microfluidic culture technologies and techniques have enabled the advancement of human disease modelling in in vitro (van Duinen et al. 2015) implementing human derived cell lines onto these chips. In 2019, 3D proximal tubule cell structures, referred to as tubuloids, were successfully cultured on chips . Following investigation for tight junctions via IF, and barrier integrity assay via the assessment of 20 kDa fluorescein isothiocyanate-(FITC-) dextran and 155kDA tetramethylrhodamine isothiocyanate-(TRITC-)dextran, tubuloid chips were capable of forming leak-tight, polarized renal tubules .
Distal tubule cells have also successfully been grown on chip technologies (Baudoin et al. 2008;Jang and Suh 2010), although there currently are few examples of human cell lines for this application. Distal tubule chips have also been used to study the effects of the pseudorabies virus on the Na + Cl − cotransporter and Na + K + -ATPase electrolyte transport mechanisms (Wang et al. 2019).

Chips for kidney drug screening
Drugs found to be successful in animal models have often failed when translated into human clinical trials (Metzner et al. 2020), highlighting the importance and impact of patient derived models. Co-cultures of renal proximal tubule-on-a-chip and human umbilical vein endothelial cells were successfully cultured under perfused conditions when grown on the OrganoPlate 3-Lane system (Vormann et al. 2021). Following exposure to cisplatin, tobramycin, and CysA, damage was observed in the presence of high concentrations of cisplatin and tobramycin but not CysA (Vormann et al. 2021). This coculture was also used to investigate ischemic injury in renal tissues.
Another example of how OOAC technology has benefitted renal drug screening was reported in 2013. Polydimethylsiloxane (PDMS) microfluidic devices containing a porous polyester membrane (pore size 0.4 μm) coated with ECM protein and collagen IV was seeded with primary human proximal tubular epithelial cells under static conditions to allow for cell adhesion (Jang et al. 2013). After 3 days, perfusion was applied to the channel containing cells (0.2 dyne/cm 2 ) to mimic conditions within the body. The bottom channel was filled with media, mimicking the interstitial space within the kidney with cisplatin (100 μM) administered via injection to the interstitial component of the chip model (Jang et al. 2013). Following exposure to cisplatin, a dramatic increase in lactase dehydrogenase (LDH) was observed in samples maintained under static conditions, while though under fluidic conditions released less LDH (Jang et al. 2013). Furthermore, after the addition of the OCT-2 inhibitor cimetidine (1 mM), cells under perfused conditions showed an almost complete prevention of cisplatin induced injury (Jang et al. 2013). This study highlights how beneficial OOAC technology has been to the study and development of renal therapies, with findings by this group like observations from the clinic.

Discussion
The use of human derived cell lines has undoubtably progressed understanding of both kidney diseases and renal specific ciliopathies. Traditional monolayer cellular models have long been the gold standard in the study of many kidney conditions, including ciliopathies, via fibroblast or hUREC models. However, 2D models have fallen short where more complex disease associated mechanisms are concerned. Furthermore, the lack of 3D structure can be especially limiting when considering drug screening trials. Recent strides have been made in renal organoid and spheroids. These 3D structures have allowed researchers to gain a deeper understanding of ciliary phenotypes, drug sensitivities and metabolism, as well as developmental processes of diseased kidneys. Traditional 2D culture models currently do not allow for the study of more complex interactions, for example the toxic effect of drugs on specific nephron specific regions.
However, like their 2D counterparts however, organoid and spheroid models have their own associated limitations including increased variability between experiments, and the formation of necrotic centers. Another, often overlooked, limitation of 3D models is the lack of vascularization. Without this vascularized network, organoids and spheroids are limited in their growth. In renal toxicity studies, it is also to consider the translational impacts this lack of vascularization may have in relation to drug delivery, clearance, and offtarget effects. Furthermore, although improved organ specific markers are present in 3D based models, there remains around 10-20% of off-target cells present in these organoid models (Romero-Guevara et al. 2020). Finally, maturation of 3D models is also important to consider when one wishes to study adult-onset renal diseases; RNA sequencing studies have found that renal organoids are more representative of embryonic kidneys than adult (Takasato et al. 2015;Kim et al. 2017;Wu et al. 2018).
Steps to overcome the limitations of traditional human renal cell-based models has led researchers to seek out alternative culturing methods. As a result, organ-on-a-chip technology has rapidly grown in both its popularity and complexity over recent years. The ability to coculture renal cell lines under fluidic environments has allowed for the expansion in many aspects of kidney disease and its treatments. Nephron segment-specific drug toxicity studies have thoroughly benefitted from chip technology, with an abundance of studies found in the current literature. Along with pre-existing humanized models, including 2D cell cultures, spheroids, and organoids, kidney-on-a-chip-based renal models will undoubtedly further advance understanding of kidney diseases, including renal ciliopathies, whilst also offering insight into the development of novel therapeutics.

Conflict of interest
The authors declare that they have no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.