The glomerulus is the filtration unit of the kidney. Injury to any component of this specialised structure leads to impaired filtration and eventually fibrosis and chronic kidney disease. Current two and three dimensional (2D and 3D) models that attempt to recreate structure and interplay between glomerular cells are imperfect. Most 2D models are simplistic and unrepresentative, and 3D organoid approaches are currently difficult to reproduce at scale and do not fit well with current industrial drug-screening approaches. Here we report a rapidly generated and highly reproducible 3D co-culture spheroid model (GlomSpheres), better demonstrating the specialised physical and molecular structure of a glomerulus. Co-cultured using a magnetic spheroid formation approach, conditionally immortalised (CI) human podocytes and glomerular endothelial cells (GEnCs) deposited mature, organized isoforms of collagen IV and Laminin. We demonstrate a dramatic upregulation of key podocyte (podocin, nephrin and podocalyxin) and GEnC (pecam-1) markers. Electron microscopy revealed podocyte foot process interdigitation and endothelial vessel formation. Incubation with pro-fibrotic agents (TGF-β1, Adriamycin) induced extracellular matrix (ECM) dysregulation and podocyte loss, which were attenuated by the anti-fibrotic agent Nintedanib. Incubation with plasma from patients with kidney disease induced acute podocyte loss and ECM dysregulation relative to patient matched remission plasma, and Nintedanib reduced podocyte loss. Finally, we developed a rapid imaging approach to demonstrate the model’s usefulness in higher throughput pharmaceutical screening. GlomSpheres therefore represent a robust, scalable, replacement for 2D in vitro glomerular disease models.
Cell-based assays are an essential element of disease modelling and drug discovery. 2D cultures do not necessarily reflect the complex microenvironment cells encounter in a tissue, in particular the interconnections between different cell types and the extracellular matrix (ECM) surrounding them. Much of our knowledge comes from studies of the interaction of tumour cells with ECM, showing that adhesion and signalling is intimately affected by these interactions, leading for example to heterogeneous drug responses1. Within a tissue, molecular concentration gradients affect various cell behaviours, including motility, migration, and signalling, which cannot be replicated in 2D models2.
Despite the progress made in advanced in vitro models, the early stages of drug development are still reliant on higher throughput 2D cell models. Utilising more advanced models only for later stages is insufficient, as false positives are compounded at each step of the drug discovery pipeline3. This highlights the importance of refining the earliest stages with better models, so that the right therapeutic candidates reach more advanced models.
3D models can be divided into scaffold-based or non-scaffold based technologies, the former being less adaptable for miniaturisation and/or high throughput screening. Non-scaffold technologies are rapidly advancing, relying on self-aggregation of cells in specialised environments4. To date these have mostly been used for tumour cells or have incorporated stem cells5. Co-culture models derived from mature cells, which represent a direct and rapid route towards a more differentiated physiological tissue structure, are lacking.
The glomerulus is a complex filter that allows the retention of cells and high molecular weight proteins in the blood, whilst allowing passage of small molecules, and electrolytes into the primary filtrate. The cell types making up the glomerular filtration barrier are glomerular endothelial cells and podocytes, on either side of a specialised basement membrane.
Conditionally immortalised cell lines of human podocytes6, GEnCs7 and mesangial cells8 are the current gold standard in vitro tools and represented a step change in differentiation profile of cultured glomerular cells. These cells are however routinely mono-cultured in 2D environments, which (for podocytes at least) has resulted in some degree of phenotypic drift9. The formation of kidney organoids from induced-pluripotent stem cells is proving to be a promising step in the right direction for in vitro models10,11. It has since been shown that it is possible to derive glomeruli and primary podocytes from these organoids which show potential for toxicity screening12. These latter models are currently limited by lengthy protocols to generate the models, and phenotypic variability between experiments.
There is, therefore, an urgent need for better, consistent in vitro models of glomerular disease, to test specific injuries and their downstream consequences. A key disease example is of immune-mediated forms of nephrotic syndrome (NS), whereby a putative circulating factor directly damages the podocyte, leading to detachment, effacement and changes to the ECM, and to date has been modelled in vitro with human podocyte 2D cultures13,14.
The intricacy of glomerular disease is difficult to model in vitro and much of our understanding thus far has been established by animal models and studies of human biopsy tissue. In vivo models of disease are limited by throughput, making them unsuitable for the early stages of drug development. For this reason, reductionist 2D cell models of disease are employed in pre-clinical testing, whereby they are used to narrow down libraries of >5000 compounds15. Many of these models are monoculture, and thus fail to replicate the intricate crosstalk environment of the glomerular filtration barrier that is understood to be important in glomerular disease progression16,17.
Here we describe a scalable spheroid model of glomerular disease (GlomSpheres), capable of modelling 3D ECM dysregulation, podocyte loss and upscaling for automated compound screening. By using magnetic spheroid formation to co-culture conditionally immortalised human podocytes and GEnCs, cells deposit an organised 3D matrix of collagen IV, laminin and fibronectin between them. Moreover, GlomSpheres exhibit an ECM switch, whereby mature collagen IVα3 and laminin α5 are deposited. Biochemical analysis revealed that GlomSphere culture massively upregulates expression of characteristic markers for podocytes (nephrin, podocin and podocalyxin) and GEnC (pecam-1) when compared to 2D cultured counterparts.
Morphologically, scanning and transmission electron microscopy (SEM and TEM) demonstrate evidence of podocyte foot-process and basement membrane formation. Confocal imaging reveals a (PECAM-1 positive) endothelial vessel-like network, surrounded by podocin and nephrin-expressing podocytes. Incubation with fibrotic agents (TGF-β1/Adriamycin) is shown to induce ECM protein (collagen IV) dysregulation and podocyte loss, which are attenuated by the addition of a commercially available anti-fibrotic agent (Nintedanib). Incubation with NS relapse and remission patient plasma, modelling direct podocyte injury, demonstrates a loss of podocytes and increase of collagen IV deposition in relapse conditions relative to remission. Co-incubation with Nintedanib attenuates this podocyte loss but not collagen IV deposition, permitting dissection of disease mechanisms.
Finally, we developed a medium/high throughput method measuring podocyte integrity in a GlomSphere model of a human genetic mutation and demonstrate rescue by chemical chaperone compounds that can be analysed rapidly in a 96-well plate.
As such, GlomSpheres represent a robust, scalable replacement for 2D in vitro glomerular disease models with much improved disease homology.
Formation and self-organisation of GlomSpheres
GlomSpheres are generated by forming a core of 5000 GEnCs, which is then wrapped with a peripheral coating of 5000 podocytes (Supplementary Fig. 1a). Over a 10-day differentiation the wrapping of peripheral podocytes gradually migrates to more completely enclose the GEnC core (Supplementary Fig. 1b). Preliminary experiments, whereby spheroids were formed by completely mixing GEnCs and podocytes, showed that cells self-organise over 72 h (Supplementary Fig. 2). The reorganised structures have a GEnC core and peripheral podocytes. The decision was made to induce this organisation manually for reproducibility via layering spheroids (as in Supplementary Fig. 1) and differentiating for 10 days.
Deposition of organised ECM proteins
Immunofluorescent staining of paraffin embedded spheroid sections revealed a basement membrane-like layer of collagen IV, which is the primary constituent of the GBM (Fig. 1c). This staining is present extracellularly in both cell types but is primarily located at the interface between the two, as would be the GBM in vivo (Fig. 1a–d). When spheroids are wholemount stained, collagen IV organisation is more difficult to visualise, but podocyte organisation at the surface is visible (Fig. 1e). Staining of fibronectin reveals it to be most present at the podocyte/GEnC interface, although it is highly expressed extracellularly by both cell types (Supplementary Fig. 3).
Switching to mature GBM proteins
Wholemount IF staining (which compared mono-cultured podocytes and GEnCs to their co-cultured, GlomSphere counterparts) examined the presence of mature and immature GBM proteins (Fig. 2). It is apparent that the immature collagen IV α1 chain was similarly expressed in 2D podocyte, GEnC and co-culture conditions (Fig. 2a, d, g, j). Whilst western blot data suggests a slight increase in collagen IV α1 expression in 2D co-culture conditions, this was not statistically significant (one-way ANOVA P = 0.2209) (Fig. 2m). Conversely, the mature collagen IV chain α3 is differentially expressed in podocyte, GEnC and GlomSpheres (Fig. 2b, e, h, k, n). Expression in podocytes and GEnC was shown to be low, with GEnC expression being marginally higher (Tukey’s multiple comparison *P = 0.0408). GlomSphere α3 expression is significantly higher than both monocultured counterparts (~23 ± 4 and ~24 ± 3 Vs ~44 ± 7) (Tukey’s multiple comparison ****P = < 0.0005). These findings were validated with western blots (Fig. 2n), where no significant difference between monoculture conditions was found but GlomSphere α3 expression was found to be significantly higher than podocyte and GEnC spheroids (~16421 ± 2475 VS ~360 ± 186 and ~332 ± 92) (Tukey’s multiple comparison ****P = < 0.0005). Similarly, the mature α5 subunit of the GBM protein laminin is shown to be absent in mono-cultured podocyte and GEnC conditions but deposited as an organised matrix in GlomSphere (Fig. 2c, f, i, l). Fluorescence intensity quantification showed Laminin α5 expression to be significantly higher in GlomSpheres than in both monocultured conditions (~47 ± 8 Vs ~11 ± .5 and ~14 ± 2) (Tukey’s multiple comparison ****P = < 0.0005). This antibody is not validated for western blot. Uncropped blots in Supplementary Fig. 7.
SEM and TEM
SEM shows GlomSphere surface detail (Fig. 3a–c). The whole GlomSphere image (Fig. 3a) demonstrated the presence of vessel-like protrusions (yellow arrows). These structures appear occasionally from GlomSpheres and stain positive for PECAM-1 (Supplementary Fig. 4). Higher power images (Fig. 3b) reveal what appear to be podocytes interdigitating (yellow arrows). Higher power images of vessel-like protrusions (yellow arrow) show what appear to be podocytes (red arrows) attached with foot processes to the vessel surface (Fig. 3c). TEM of a section through a GlomSphere reveals interdigitation of podocytes at the edges (Fig. 3d). This is further observed in Fig. 3e, f, as well as the presence of fibrillar collagen IV (yellow arrows). A tri-layer structure is shown in Fig. 3g, whereby fibrillar collagen (yellow arrows) is deposited in the extracellular space between a neighbouring podocyte and endothelial cell.
Upregulation of cell-specific marker proteins
The expression of key podocyte and GEnC proteins in 2D and GlomSphere samples was examined with human glomerular lysate as a positive control. Western blotting for podocalyxin (which coats podocyte secondary foot-processes) and PECAM-1 (which is located at endothelial cell–cell junctions) was performed (Fig. 4a, b). Whilst podocalyxin appears to be absent in the 2D lysate, it is significantly upregulated in GlomSpheres (~6 fold ± 0.3) (Tukey’s multiple comparison ****P = < 0.0001) (Fig. 4a). Compared to the glomerular control, podocalyxin is lower in GlomSpheres (~10 fold increase in Glomerulus VS ~6 fold increase in GlomSphere) (Tukey’s multiple comparison ****P = < 0.0001).
Likewise, whilst PECAM-1 appears to be weakly present in 2D lysates, its expression in GlomSpheres is significantly increased and is approaching the expression level of glomerular lysate (~2.5 fold ± 0.5) (Tukey’s multiple comparison ***P = 0.0006) (Fig. 4b). The presence of Podocalyxin and PECAM-1 in GlomSpheres was validated using IF staining techniques. Podocalyxin (green) is shown to be present in GlomSpheres and is located primarily at the outer edges, where podocytes are located (Fig. 4c). Under the same illumination settings, podocalyxin staining is negligible in 2D cultured cells (Fig. 4e). Conversely, pecam-1 expression (red) is shown to be present in GlomSpheres but is localised at the core, where the GEnCs are located (Fig. 4d). Staining for PECAM-1 was weakly present in 2D co-cultured cells under the same illumination. Collectively, these results indicate that 2D culture results in diminished podocyte and GEnC marker expression and that GlomSphere culture is restorative. Uncropped blots in Supplementary Fig. 7.
The presence of key podocyte markers podocin and nephrin was then examined. Wholemount triple-IF staining for PECAM-1, nephrin and collagen IV (Fig. 5a–d) demonstrates an organised central core of PECAM-1 staining, surrounded by a mostly peripheral nephrin stain. collagen IV staining is concentrated primarily at the area between PECAM-1 and nephrin. This staining pattern is repeated for the PECAM-1, podocin and collagen IV triple stain (Fig. 5e–h), with podocin staining located peripherally. The intricate, vessel-like organisation of PECAM-1 staining can be more clearly observed (in Supplementary Movie 1), whereby a branching vascular network is at the core of GlomSpheres.
QPCR was employed to examine nephrin and podocin mRNA levels (Fig. 5i, j, respectively). It is apparent that whilst there is minimal expression and very little difference in nephrin mRNA between each of the 2D culture conditions, there is a large and significant increase when podocytes and GEnCs are cultured as GlomSpheres (600 fold increase, Tukey’s multiple comparison **P = 0.0092) (Fig. 5i). The same is true for podocin mRNA, which is also not significantly different between 2D cultures, but there is an even greater increase in GlomSphere cultures (30,000 fold increase, Tukey’s multiple comparison ****P = < 0.0001) (Fig. 5j). A comparison of GlomSphere and human kidney cortex expression of podocin and nephrin was also made as a positive control for primers (Supplementary Fig. 6). Despite variability between cortex samples (likely due to differing glomeruli counts), expression levels were stronger than GlomSpheres (more so for nephrin than podocin, which was closer to in vivo expression).
Induction and attenuation of disease phenotypes in GlomSpheres
A series of experiments was then undertaken to manipulate disease phenotypes in GlomSpheres. Initially, GlomSpheres were cultured in the presence of profibrotic agent TGF-β1, with or without Adriamycin, a reno-toxic agent known to induce a focal segmental glomerulosclerosis (FSGS) phenotype in rat models18,19 (Fig. 6a–f). GlomSphere incubation with 10 ng/ml TGF-β1 and 10 nM Adriamycin for 72 h was shown to induce dramatic podocyte detachment from the spheroid’s surface compared to untreated GlomSpheres (Fig. 6a, b). A clarification of podocyte detachment is shown in Supplementary Fig. 5. Quantification of podocyte (GFP) mean fluorescence intensity confirms a statistically significant loss of podocyte fluorescence after TGF-β1 and Adriamycin treatment (~87 ± 9 Vs ~19 ± 3, respectively) (Tukey’s multiple comparison ****P = < 0.0005) (Fig. 6d). This loss is also observed when quantifying podocyte percentage coverage (~41% Vs ~2.5%, respectively). Nintedanib is an inhibitor of platelet-derived growth factor receptor, fibroblast growth factor receptor and vascular endothelial growth factor and is a licenced treatment for idiopathic pulmonary fibrosis (IPF) with antifibrotic properties20,21,22. The addition of 5 nM Nintedanib at the same time as the TGF-β1 and Adriamycin treatment reduced the loss of podocytes from the surface of the spheroid (Fig. 6c). Quantification of podocyte (GFP) mean fluorescence intensity and podocyte coverage confirms a statistically significant retention of podocyte fluorescence compared to TGF-β1 and Adriamycin only (~19 ± 3 TGF-β1 and Adriamycin Vs ~39 ± 6 with the addition of Nintedanib) (Tukey’s multiple comparison ****P = < 0.0005) (Fig. 6d) and (~2.5% Vs ~10.5%) (Fig. 6e). Interestingly, collagen IV fluorescence appears to have been lost in the TGF-β1/Adriamycin and Nintedanib conditions (Fig. 6f).
These findings led to the hypothesis that Adriamycin-mediated podocyte loss was occurring before excessive production of collagen IV could occur. To test this, TGF-β1 (without Adriamycin) was administered to GlomSpheres for 72 h (Fig. 6g–i). It is apparent that TGF-β1 incubation induced a dramatic increase in collagen IV deposition when compared to the untreated control (Fig. 6g, h). This is confirmed by quantification of collagen IV mean fluorescence intensity, which shows a statistically significant increase in fluorescence after TGF-β1 treatment (~36 ± 5 Vs ~69 ± 22 Tukey’s multiple comparison ***P = 0.0003) (Fig. 6h). This increase in collagen IV fluorescence appears to have been attenuated with the addition of Nintedanib (~69 ± 22 Vs 38 ± 8) (Fig. 6i, j) (Tukey’s multiple comparison ***P = 0.0006). Interestingly, podocyte loss appears to have been negligible across all three conditions (Fig. 6g, h, i, k, l). Jointly, these experiments highlight two capabilities of the GlomSphere model, one of which models early podocyte loss (TGF-β1 and Adriamycin), the other modelling fibrotic ECM dysregulation (TGF-β1 only).
GlomSphere incubation with nephrotic syndrome (NS) patient plasma
Whilst an artificially induced disease phenotype is a powerful tool, a greater test of a GlomSphere’s ability to model glomerular disease is to examine its response to nephrotic syndrome (NS) patient plasma. The samples were from plasma exchange effluent of patients suffering immediate recurrence of their disease post-renal transplantation (with institutional ethical permissions)14. GlomSpheres were incubated with either NS relapse, remission, or relapse + Nintenadib human plasma samples for 7 days (Fig. 7). Podocyte loss was first examined (Fig. 7a–c, g, h). It is apparent that podocyte coverage is weaker in relapse GlomSpheres (Fig. 7a) relative to remission (Fig. 7b) and relapse + Nintedanib conditions (Fig. 7c). Quantification of GFP fluorescence intensity and podocyte coverage confirms that relapse plasma significantly reduces podocyte attachment (intensity ~9 ± 1 Vs ~17 ± 2) (Tukey’s multiple comparison ****P = < 0.0001) (podocyte coverage ~29% Vs ~57% (Tukey’s multiple comparison ****P = < 0.0001). It is also clear that the addition of Nintedanib attenuates this loss significantly when compared to relapse only (intensity ~9 ± 1 Relapse Vs 13 ± 2 Relapse + Nintedanib) (Tukey’s multiple comparison ****P = < 0.0001) (podocyte coverage ~29% Relapse Vs ~46% Relapse + Nintedanib) (Tukey’s multiple comparison ****P = < 0.0001). Collagen IV deposition also appears to increase in relapse conditions, relative to remission (Fig. 7d, e). Relapse + Nintedanib collagen IV appears to have a different staining pattern (Fig. 7d, f). Quantification of collagen IV mean fluorescence intensity confirms that relapse is significantly higher than remission (~100 ± 20 Vs ~61 ± 15 Tukey’s multiple comparison (**P = < 0.0014). Collagen IV fluorescence is shown to be significantly higher in the relapse + Nintenadib condition relative to both relapse and remission (~129 ± 51 Vs ~100 ± 20 and ~61 ± 15, respectively, Tukey’s multiple comparison *P = 0.0401 and ****P = < 0.0001, respectively). This indicates that whilst Nintedanib appears to change the collagen IV deposition pattern, it is not able to attenuate dysregulation and may exacerbate it.
A rapid pharmaceutical screening approach
The nature of spheroid culture lends itself well to higher throughput screens. To demonstrate the ability of GlomSpheres to run a rapid pharmacological screen, a rapid imaging approach was employed using the In-cell analyser. A podocyte cell-line generated from a patient with a pathogenic R138Q mutation of podocin14 was used for these experiments to study the ability of various compounds to rescue podocin functionality. Podocyte loss from the spheroid was quantified as a surrogate for podocyte adhesiveness (which is a well-established readout of podocyte pathology in in vitro assays) and was compared in wildtype (WT), podocin mutant (PM) and PM cells treated with various compounds. These compounds are derived from a previously screened NCBI library, where a single compound (termed 407) has been found to rescue podocin functionality both in vitro and in vivo. The current experiment tested this compound against chemically-modified variants to assess relative podocyte phenotype rescue efficacy. After 5 days of treatment, podocyte attachment (via GFP expression) was compared (Fig. 8).
Firstly, it is shown that wildtype podocytes (Fig. 8b) attach more strongly than podocin mutant podocytes (Fig. 8a). This is confirmed by quantification of GFP mean fluorescence intensity within spheroid area (Fig. 8d) and podocyte coverage percentage (Fig. 8e). Secondly, the ability of a suite of compounds to restore podocyte attachment to that of wildtype podocytes is also shown (one of which is shown in Fig. 8c). This is confirmed through quantifications, which also demonstrate the ineffectiveness of other compounds (Fig. 8d, e).
A standard podocyte adhesion assay was performed to compare GlomSphere results to a more traditional 2D, monoculture assay of cell attachment (Fig. 8f). These results suggest the GlomSphere assay has greater dynamic range than a 2D assay. The broad direction of results between the two assays is similar for each compound (with some notable exceptions such as 606) but the range of differences is extended in the 3D assay, with more compounds in the latter therefore reaching statistical significance.
The data presented here demonstrate an effective method of enhancing the phenotype of conditionally immortalised human podocytes and GEnCs in 3D. It is apparent that the superior environment presented by spheroid culture induces an organised, glomerular basement membrane (GBM)-like extracellular matrix at the podocyte/GEnC interface. Importantly, we have demonstrated that by co-culturing podocytes and GEnCS in spheroids, mature collagen IV α3 and laminin α5 GBM proteins are dramatically upregulated. This has never previously been achieved in 2D culture models23. This is reminiscent of the developmental switch that happens in vivo, whereby glomerular isoforms of collagen IV transition from predominantly immature COL4α1α2α1 networks to mature COL4α3α4α5 networks24. Likewise, during glomerulogenesis laminin-111 α1β1γ1 is replaced by mature laminin-521 α5β2γ125.
The interdigitating foot processes seen in TEM and SEM images (Fig. 3) are further indicative of a glomerulus-like maturation of podocytes in GlomSpheres. The ability of podocytes to interdigitate under GlomSphere culture may also be underpinned by the upregulation of the critical slit diaphragm proteins nephrin and podocin (Fig. 5) and suggest more physiological slit diaphragm signalling.
The vessel-like protrusions seen in Fig. 3 are suggestive of glomerulus like maturation in GEnCs, especially as they appear to be PECAM-1 positive (Supplementary Fig. 4). Figure 5 and Supplementary Movie 1 demonstrate that this network of PECAM-1 expressing vessels is present throughout the centre of GlomSpheres and Fig. 4 indicates that whilst PECAM-1 expression is extremely low in 2D cultured cells, in GlomSpheres it is returned to almost glomerular levels. These results are interesting when compared to the findings of Rinschen et al.9 which demonstrated low levels of glomerular marker proteins from 2D monocultured conditionally immortalised (CI) podocytes. Our data suggest that 3D co-culture of CI GEnCs and podocytes as part of GlomSpheres may have a restorative effect.
Experiments that artificially modulate disease phenotypes in GlomSpheres display their potential for pharmaceutical screening. TGFβ1 and Adriamycin are known drivers of fibrosis26,27,28 and as hypothesized, their combined application induced podocyte loss in GlomSpheres (Fig. 6). Nintedanib has previously been shown to interfere with fibroblast proliferation and differentiation and extracellular matrix deposition29,30. It was hypothesized that the addition of Nintedanib to the fibrotic milieu would attenuate the induced fibrotic phenotype somewhat and this is apparent in Fig. 6, where podocyte loss is reduced with Nintedanib addition.
Adriamycin is known to induce ECM remodelling but also highly toxic to cells31,32. It was hypothesized that podocytes were being lost from GlomSpheres before ECM remodelling could occur. TGF-β1 was therefore applied to GlomSpheres as the sole fibrotic stimulus and as predicted, ECM dysregulation was observed (Fig. 6). This dysregulation is also attenuated by the antifibrotic properties of Nintedanib. Taken together, these fibrotic modulation experiments demonstrate the ability of GlomSpheres to model two key aspects of glomerulosclerosis, podocyte loss and ECM dysregulation, and to examine them separately. Moreover, the demonstrated attenuation properties of Nintedanib demonstrate GlomSphere’s potential as platform for drug-screening.
Incubation with NS patient plasma samples allows for an assessment of GlomSphere’s response to a disease milieu, which will contain disease-altering circulating factors other than TGF-β114. As hypothesized, FSGS relapse plasma was shown to induce both podocyte loss and an increase in collagen IV deposition, relative to remission plasma (Fig. 7), which is distinct from the effects of TGFβ1or Adriamycin, in this system. Nintedanib appears to have a similar protective effect on podocyte loss in FSGS relapse plasma to that observed with TGFβ1/Adriamycin conditions (Fig. 6 vs Fig. 7). This protective effect is however not observed with regards to collagen IV deposition, highlighting the limits of Nintedanib treatment in this case. Whilst it would be optimal to compare relapse and remission patients to a normal control patient, healthy patients do not undergo plasmapheresis like FSGS patients. For this reason, the internal control offered by examining the same patient as different dates (in relapse and remission) is the fairest control.
The use of the In cell analyser (Fig. 8) to conduct a higher throughput drug screen demonstrates the rapidity and scalability of GlomSpheres. This assay was performed on live cells and imaging an entire 96-well plate takes <5 min. Quantification can also be performed automatically on the system, removing human error and increasing throughput. This experiment was limited to 12–18 spheroids per condition due to limited quantities of precious compounds. There is however no reason an assay of this nature cannot be upscaled considerably. Interestingly, comparison with the 2D adhesion assay shows the greater dynamic range of the GlomSphere assay. One of these was the compound termed “407”, which all other compounds are variants of. This compound has been shown to be effective in animal studies33 and its increased potency in the GlomSphere assay suggests increased physiological relevance of this model.
There are limitations with the GlomSphere model, including the lack of glomerular mesangial cells, which can contribute to the signalling milieu as well as mesangial derived matrix components, and structural support for capillaries34. The latter is unlikely to be adequately recreated unless flow is also provided, which is currently beyond the capabilities of this type of model, though it was interesting that primitive vascularisation is seen in GlomSpheres, indicating that enough soluble signals (e.g. VEGF) are present to initiate capillary formation. The GlomSphere model is also currently not a model of glomerular function (i.e. does not model filtration efficiency). We are currently exploring possibilities to modify the model to this end, although scalability is likely to be compromised in this case.
3D culture is emerging as an essential tool to span the gap between 2D culture and animal models, with a far greater ability to mimic the behaviour of in vivo tissue. Practical limitations to date include the need to incorporate pluripotent stem cells (in some systems) or external matrices, reproducibility, expense, and complexity particularly if requiring technical scaffolds. These shortcomings have resulted in very little change to the industrial drug-discovery pipeline, which is still reliant on over-simplistic 2D cell models. We have developed technique that overcomes both the problem of representativeness and scalability/reproducibility, as well as being extremely straightforward to analyse.
Collectively, our data demonstrate that GlomSpheres represent an excellent method of rapidly and reproducibly 3D co-culturing glomerular cell lines. More broadly this method shows the potential to co-culture mature differentiated cells in a way that more precisely recreates 3D tissue properties. There is an urgent need for effective treatments for glomerular diseases, and for antifibrotic drugs, which are more likely to be discovered using a model that accurately simulates fibrosis. To the best of our knowledge, GlomSpheres represent a new in vitro standard, as a scalable, 3D model of the glomerulus with great potential as a model for drug-screening.
Cell lines and cell culture
Conditionally immortalised podocytes and GEnCs were isolated and cultured as previously described6,7. When required, these cell lines were further transfected to generate stable GFP-actin overexpressing podocytes and mCherry-actin overexpressing GEnCs as previously described34. Podocin-mutant (PM) cells were used as previously described14. Each individual cell line is from a different single patient, and as such, are from different genetic backgrounds. Several clones of each are taken and used interchangeably in our lab. No difference between clones is observed. For all co-culture experiments, cells were cultured in EBM-basal culture medium (Lonza) supplemented with Bulletkit growth factors to produce EGM-2 MV growth medium (Lonza). During proliferation, cells were cultured at 33 °C, before being differentiated at 37 °C for 10 days. ‘2D co-culture’ conditions in experiments are achieved by mixing podocytes and GEnCs at a 1:1 ratio before seeding on tissue culture plastic.
GlomSphere cell culture
Spheroids were formed using a modified version of N3D bioscience standard protocol for magnetic spheroid bioprinting (Nano3D Biosciences Inc)(Greiner 655840). A T75 flask of cells was incubated overnight with 100 µl nanoshuttle-PL (Nano3D Biosciences Inc)(Greiner 657841), added to 10 ml of fresh cell culture medium. Cells were then washed with 5 ml sterile phosphate-buffered saline (PBS) and with trypsin-EDTA (Lonza). Cells were then pelleted via 1500 rpm centrifugation (5 mins) and counted with a Luna cell counter (Logos Biosystems). For monoculture experiments, 10,000 cells were pipetted into each well of an ultra-low attachment plate (Greiner) and a 96-magnet MagDrive (Nano3D Biosciences) was place underneath. Spheroids were left to form overnight at 33 °C.
For co-culture/GlomSphere spheroids, a spheroid of 5000 GEnCs was generated as above and allowed to aggregate for 1 h at 33 °C. The MagDrive was then removed and 5000 podocytes were pipetted into each well before the MagDrive was replaced. This forces the newly added podocytes to form a peripheral coating around the GEnC core. The spheroids were then thermoswitched to 37 °C and the podocyte layer migrates (Supplementary Fig. 1). After 10 days of differentiation, treatment or fixation was performed. Magnetic beads stay attached to the outer membrane of cells for the length of the experiment and are inert and non-fluorescent.
Treatments with fibrotic and anti-fibrotic agents
Under treatment conditions, GlomSphere cell culture media was supplemented with 10 ng/ml TGF-β1 (R&D systems) and 10 µM Adriamycin (Sigma D1515) for 72 h. When stated, 10 µM Nintedanib was added at the same time as the fibrotic agent(s).
Whole-mount fixation and immunofluorescent staining
GlomSpheres were magnetically transferred to Eppendorf tubes and fixed in 4% Paraformaldehyde (Sigma) containing 1% Triton-x100 (Sigma) (20 mins, 20 °C). To block, spheroids were incubated with 5% bovine-serum-albumin (BSA) (Sigma) containing 0.1% Triton-x100 (overnight, 4 °C). Spheroids were then incubated with primary antibody, diluted in 5% BSA containing 0.1% Triton-x100 (48 h, 4 °C). Spheroids were then washed in PBS containing 1% Triton-x100 (3 × 30 min, room temp). Spheroids were then incubated with secondary fluorescent antibodies diluted 1:400 in 5% donkey serum (Sigma) containing 0.1% Triton-x100 (24 h, 4 °C). Spheroids were then washed in PBS containing 1% Triton-x100 (3 × 30 min, 20 °C).
Sectioning and 2D immunofluorescent staining
Sections were cut as previously described34. Sections or fixed 2D-cultured cells on glass coverslips were washed in PBS (2 × 5 mins, 20 °C). To block, samples were incubated in 5% BSA containing 0.1% Triton-x100 (45 min, 20 °C), then incubated with primary antibody, diluted in 5% BSA containing 0.1% Triton-x100 (1 h, 20 °C). Primary antibodies were then washed with PBS (2 × 5 mins, 20 °C) before being incubated with fluorescent secondary antibodies diluted in 1:400 in 5% donkey serum (1 h, 20 °C). Samples were then washed with PBS (2 × 5 mins, 20 °C) and mounted to microscope slides with 5 µl Mowiol (Sigma).
Mid/high throughput screen
Spheroids were formed in 96-well plates, using either GFP CI podocytes (WT) or GFP CI podocin mutant podocytes (PM). Compounds were added 24 h after culture and re-dosed daily. Images were taken daily and those shown and quantified are from day 5. Imaging was performed using an In-cell analyser (GE lifesciences).
Collagen IV pan (Abcam ab6586) (1:200), Col IVα1 and α3 (Chondrex 7070 and 7076, respectively) (1:100), Laminin α5 (Abcam ab77175) (1:200), fibronectin (Abcam ab2413) (1:50), podocin (Abcam ab50339) (0.2 µg/ml), nephrin (Origene BP5030)(1:50), PECAM-1 (Cell Signalling #3528)(1:800), podocalyxin (ThermoFisher 39-3800)(1:2000).
Two 96-well plates of spheroids were pooled (192 spheroids). Spheroids were disintegrated by mechanical force using a syringe and RIPA buffer (ThermoFisher) Standard western blots were then performed as previously described34.
Cells are suspended in media at a concentration of 3 × 105/ml and incubated upright in a tube (10 min, 37 °C). Cells are added to the wells of a 96-well plate (50 µl/well) along with 50 µl of PBS and any inhibitor being tested. Leave cells to adhere (duration is cell dependent). Control wells (for 100% attachment) are then fixed in 4% PFA (20 mins, room temp). Wash all cells with (PBS 2× room temp) and fix remaining wells in 4% PFA (20 min, room temp). Wells are washed with 100 µl H2O (×3) before being stained with 0.1% crystal violet in 2% ethanol (60 mins, room temp). Cells are then washed with H2O (x3)
Two 96-well plates of spheroids were pooled (192 spheroids). Spheroids were disintegrated by mechanical force using a syringe and RNA isolated with a RNeasy kit (Qiagen). cDNA was synthesized using a high capacity RNA to cDNA kit (Thermofisher).
Podocin forward: GTCCTCGCCTCCCTGATCTT
Podocin reverse: AGGAAGCAGATGTCCCAGTC
Nephrin forward: CCGGGAGACGCCTTAAACTT
Nephrin reverse: AGCCTTTGAATGGGGCTCTC
GAPDH forward: GAAGGTCGGAGTCAACGGATT
GAPDH reverse: CGCTCCTGGAAGATGGTGAT
Imaging and analysis
Confocal light microscopy was performed using SP8 spinning disc confocal microscope attachment (Leica), attached to a DMI 6000 inverted microscope (Leica). Widefield microscopy was performed on a Leica DM IRB linked to a Zeiss Axiocam ERc 5S. Lightsheet microscopy was performed using a Zeiss Z.1 lightsheet microscope.
All image analyses and quantifications were done using ImageJ software. For quantifications of IF images, a region of interest (ROI) was drawn around the measured structure (i.e. the perimeter of a spheroid) and mean fluorescence intensity in this area was calculated. In this way, pixels within the image that are absent of fluorescent signal are included in calculation, as well as the varying intensity of pixels positive for signal. In our experience this provides a more representative quantification than subjectively setting a threshold for the image and measuring total pixels present for a signal.
For podocyte coverage percentage, images were thresholded in imageJ and the same ROI was used to measure the percentage of spheroid area positive for podocyte signal (and therefore a measure of relative podocyte adhesion to GlomSphere surface).
Statistics and reproducibility
Statistical analysis was performed using GraphPad Prism 7 software. Statistical tests used and sample sizes for each experiment are detailed in their corresponding figure legends. Unless stated otherwise, repeats are combined biological and technical repeats (that is, pooled and analysed spheroids from several experiments, each containing several spheroids).
Patient plasma samples
Patient plasma samples used are discarded plasma from patients undergoing plasmapheresis. They were obtained from patients, with ethical permission (National IRAS research ethics approval 09/H0106/80), from centres within the UK. Samples are immediately aliquoted and stored at −80 °C with freeze–thaw cycles avoided. Samples were taken from patients undergoing their first session of plasmapheresis for relapse or remission of FSGS14. Plasma is diluted in basal (non-supplemented) EBM-2 media at 15% concentration.
Muranen, T. et al. Inhibition of PI3K/mTOR leads to adaptive resistance in matrix-attached cancer cells. Cancer Cell 21, 227–239 (2012).
Langhans, S. Three-dimensional in vitro cell culture models in drug discovery and drug repositioning. Front. Pharmacol. 9, 6 (2018)
Scannell, J. W. & Bosley, J. When quality beats quantity: decision theory, drug discovery, and the reproducibility crisis. PLoS ONE. 11, e0147215 (2016).
Sutherland, R. Cell and environment interactions in tumor microregions- the multicell spheroid model. Science 240, 177–184 (1988).
Jensen, C. & Teng, Y. Is it time to start transitioning from 2D to 3D cell culture? Front. Mol. Biosci. 7, 33 (2020).
Saleem, M. A. et al. A conditionally immortalized human podocyte cell line demonstrating nephrin and podocin expression. J. Am. Soc. Nephrol. 13, 630–638 (2002).
Satchell, S. C. et al. Conditionally immortalized human glomerular endothelial cells expressing fenestrations in response to VEGF. Kidney Int 69, 1633–1640 (2006).
Sarrab, R. M. et al. Establishment of conditionally immortalized human glomerular mesangial cells in culture, with unique migratory properties. Am. J. Physiol. Ren. Physiol. 301, F1131–F1138 (2011).
Rinschen, M. M. et al. A multi-layered quantitative in vivo expression atlas of the podocyte unravels kidney disease candidate genes. Cell Rep. 23, 2495–508. (2018).
Takasato, M. et al. Kidney organoids from human iPS cells contain multiple lineages and model human nephrogenesis. Nature 526, 564–568 (2015).
Takasato, M., Er, P. X., Chiu, H. S. & Little, M. H. Generation of kidney organoids from human pluripotent stem cells. Nat. Protoc. 11, 1681–92. (2016).
Hale, L. J. et al. 3D organoid-derived human glomeruli for personalised podocyte disease modelling and drug screening. Nat. Commun. 9, 5167 (2018).
Saleem, M. Molecular stratification of idiopathic nephrotic syndrome. Nat. Rev. Nephrol. 15, 751–66 (2019).
Harris, J. et al. Active proteases in nephrotic plasma lead to a podocin-dependent phosphorylation of VASP in podocytes via protease activated receptor-1. J. Pathol. 229, 660–671 (2013).
Fu, J., Lee, K., Chuang, P. Y., Liu, Z. H. & He, J. C. Glomerular endothelial cell injury and cross talk in diabetic kidney disease. Am. J. Physiol. Renal. Physiol. 308, F287–F97. (2015).
Eremina, V. et al. VEGF inhibition and renal thrombotic microangiopathy. N. Engl. J. Med. 358, 1129–1136 (2008).
Matthews, H., Hanison, J. & Nirmalan N. “Omics”-informed drug and biomarker discovery: opportunities, challenges and future Perspectives. Proteomes 4, 28 (2016).
Sternberg, S. S. Cross-striated fibrils and other ultrastructural alterations in glomeruli of rats with daunomycin nephrosis. Lab. Invest. 23, 39–51 (1970).
Lee, V. W. & Harris, D. C. Adriamycin nephropathy: a model of focal segmental glomerulosclerosis. Nephrol. (Carlton) 16, 30–38 (2011).
Hostettler, K. E. et al. Anti-fibrotic effects of nintedanib in lung fibroblasts derived from patients with idiopathic pulmonary fibrosis. Respir. Res 15, 157 (2014).
Case, A. H. & Johnson P. Clinical use of nintedanib in patients with idiopathic pulmonary fibrosis. BMJ Open Respir. Res. 4, e000192 (2017).
Nanthakumar, C. et al. Dissecting fibrosis: therapeutic insights from the small-molecule toolbox. Nat. Rev. Drug Discov. 14, 693–720 (2015).
Lennon, R. et al. Global analysis reveals the complexity of the human glomerular extracellular matrix. J. Am. Soc. Nephrol. 25, 939–951 (2014).
Khoshnoodi, J., Pedchenko, V. & Hudson, B. G. Mammalian collagen IV. Microsc Res. Tech. 71, 357–370 (2008).
Goldberg, S., Adair-Kirk, T. L., Senior, R. M. & Miner, J. H. Maintenance of glomerular filtration barrier integrity requires laminin alpha5. J. Am. Soc. Nephrol. 21, 579–86 (2010).
Meng, X. M., Nikolic-Paterson, D. J. & Lan, H. Y. TGF-β: the master regulator of fibrosis. Nat. Rev. Nephrol. 12, 325–338 (2016).
Bucciarelli, E., Binazzi, R., Santori, P. & Vespasiani, G. [Nephrotic syndrome in rats due to adriamycin chlorhydrate]. Lav. Ist. Anat. Istol. Patol. Univ. Stud. Perugia 36, 53–69 (1976).
Chen, A. et al. Experimental focal segmental glomerulosclerosis in mice. Nephron 78, 440–452 (1998).
Wollin, L., Maillet, I., Quesniaux, V., Holweg, A. & Ryffel, B. Antifibrotic and anti-inflammatory activity of the tyrosine kinase inhibitor nintedanib in experimental models of lung fibrosis. J. Pharm. Exp. Ther. 349, 209–220 (2014).
Wollin, L. et al. Mode of action of nintedanib in the treatment of idiopathic pulmonary fibrosis. Eur. Respir. J. 45, 1434–1445 (2015).
Outomuro, D., Grana, D. R., Azzato, F. & Milei, J. Adriamycin-induced myocardial toxicity: new solutions for an old problem? Int J. Cardiol. 117, 6–15 (2007).
Injac, R. & Strukelj, B. Recent advances in protection against doxorubicin-induced toxicity. Technol. Cancer Res Treat. 7, 497–516 (2008).
Odolczyk, N. et al. Discovery of novel potent ΔF508-CFTR correctors that target the nucleotide binding domain. EMBO Mol. Med. 5, 1484–1501 (2013).
Tuffin, J. et al. A composite hydrogel scaffold permits self-organization and matrix deposition by cocultured human glomerular cells. Adv. Health. Mater. 8, e1900698 (2019).
This work was part funded by a BBSRC CASE studentship jointly with UCB pharmaceuticals (Slough). Additional funding was provided by Wellcome Trust ISSF, Medical Research Council (MR/R013942/1), the Elizabeth Blackwell Institute (TRACK) and the Nephrotic Syndrome Trust’. M.C. is funded by the Royal Thai Government Scholarship program. Confocal (BBSRC Alert 13 capital grant BB/L014181/1), Light sheet (Wellcome Trust), TEM and SEM (BBSRC 17ALERT) imaging was performed at the Wolfson Bioimaging Facility. TEM and SEM sample preparation and imaging was performed by Dr. Chris Neal and Mrs Judith Mantell.
T.J. and UCB Pharmaceuticals are using a variant of this model for drug screening. No other competing interests.
Peer review information Communications Biology thanks Ilse Daehn and the other, anonymous, reviewers for their contribution to the peer review of this work. Primary Handling Editor: Eve Rogers. Peer reviewer reports are available.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Tuffin, J., Chesor, M., Kuzmuk, V. et al. GlomSpheres as a 3D co-culture spheroid model of the kidney glomerulus for rapid drug-screening. Commun Biol 4, 1351 (2021). https://doi.org/10.1038/s42003-021-02868-7