Recapitulating kidney development in vitro by priming and differentiating mouse embryonic stem cells in monolayers

In order to harness the potential of pluripotent stem cells, we need to understand how to differentiate them to our target cell types. Here, we developed a protocol to differentiate mouse embryonic stem cells (ESCs) to renal progenitors in a step-wise manner. Microarrays were used to track the transcriptional changes at each stage of differentiation and we observed that genes associated with metanephros, ureteric bud, and blood vessel development were significantly upregulated as the cells differentiated towards renal progenitors. Priming the ESCs and optimizing seeding cell density and growth factor concentrations helped improve differentiation efficiency. Organoids were used to determine the developmental potential of the renal progenitor cells. Aggregated renal progenitors gave rise to organoids consisting of LTL+/E-cadherin+ proximal tubules, cytokeratin+ ureteric bud-derived tubules, and extracellular matrix proteins secreted by the cells themselves. Over-expression of key kidney developmental genes, Pax2, Six1, Eya1, and Hox11 paralogs, during differentiation did not improve differentiation efficiency. Altogether, we developed a protocol to differentiate mouse ESCs in a manner that recapitulates embryonic kidney development and showed that precise gene regulation is essential for proper differentiation to occur.


Pluripotent stem cells (PSCs), such as embryonic stem cells (ESCs)
and induced PSCs (iPSCs), have the ability to self-renew and differentiate into all cells of all three germ layers, including renal progenitors. Renal progenitors, which include nephron progenitors, ureteric bud cells, and stromal progenitors, are multipotent stem cells capable of self-renewal and differentiation to all the specialized cell types in the adult kidney. These properties make renal progenitors a valuable cell source for therapeutic applications and regenerative medicine. Since the renal progenitor pool is exhausted shortly after birth, we propose to produce embryonic kidney-like cells in vitro by differentiating PSCs to multipotent renal progenitors. When cultured under conditions that mimic embryonic kidney development, PSCs can be differentiated to renal progenitors and coaxed to form kidney organoids, and these cells can be used in various applications, including cell-based therapy and nephrotoxicity screening. Here, we improved on existing mouse differentiation protocols by eliminating the use of embryoid bodies and developing a monolayer protocol that generates multiple renal progenitor cell types, including nephron progenitors, ureteric bud cells, and stromal progenitors.
The strategy to developing an in vitro differentiation method is to recapitulate embryonic kidney development. Mammalian kidney development begins with the formation of the pronephric duct at around E8.0 in mice, which extends caudally towards the hind limbs. As the nephric duct extends, it induces the formation of the mesonephros on the adjacent intermediate mesoderm at around E9.0 in mice. The metanephros starts to form at around E10.5 in mice when the nephric duct reaches a mass of cells called metanephric blastema. The nephric duct induces the metanephric blastema to become the metanephric mesenchyme, and, in turn, the metanephric mesenchyme induces the formation and outgrowth of the ureteric bud from the nephric duct. The cross-talk between the metanephric mesenchyme and ureteric bud is important for survival, migration, and proliferation of the two cell populations. The metanephric mesenchyme condenses to form pretubular aggregates, and the pretubular aggregates continue to undergo morphological changes and mesenchymal-to-epithelial transition to give rise to renal vesicles, comma-shaped bodies, Sshaped bodies, and eventually connect with the collecting duct and ureter to form a continuous tubular system.
From seminal research studies on the cellular and molecular basis of kidney induction and patterning, we know that activin A, bone morphogenetic proteins (Bmps), fibroblast growth factors (Fgfs), and Wnt molecules play important roles in inducing and promoting renal differentiation from primitive streak cells [1][2][3][4] . In addition to regulatory molecules, some key transcriptional factors and co-activators directly involved in inducing and maintaining renal progenitors include Pax2, Six1, Eya1, and Hox11 paralogs (Hoxa11, Hoxc11, and Hoxd11). Knockout mouse studies show that defects in these genes result in metanephric mesenchyme apoptosis and lack of ureteric bud formation [5][6][7][8] . In humans, defects in these genes result in various renal abnormalities, including branchio-oto-renal syndrome, renal hypodysplasia, and renal-coloboma syndrome 9 .
There has been tremendous advancement in human PSC differentiation, but very little advancement in mouse PSC differentiation. Given the importance and power of transgenic rodent models in studying kidney development and disease, we recognize the need for a more reliable, reproducible, and a better characterized murine renal differentiation protocol. Currently, almost all mouse differentiation protocols use embryoid bodies as an intermediate step to generate renal progenitors [10][11][12][13][14][15] . Embryoid bodies are formed by aggregating and culturing PSCs as a cell aggregate, and factors such as embryoid body size and cell position within the embryoid body would affect differentiation. These protocols opt to use embryoid bodies as a method of differentiation, probably because mouse PSCs are notorious for being difficult to grow as a monolayer due to their "naive" pluripotency state as opposed to the "primed" pluripotency state of human PSCs, which are easier to grow as a monolayer. In addition, these studies did not differentiate ESCs to renal progenitors in a step-wise manner to recapitulate embryonic kidney development and often rely on transplanting the embryoid bodies into mice for maturation 10,11 . Lastly, only one progenitor population is generated in these protocols, while developmental studies demonstrate that the interactions between nephron progenitors, ureteric bud cells, and stromal progenitors are crucial for proper kidney development to occur.
Understanding these issues, we developed a protocol to direct the differentiation of mouse ESC, as a monolayer, to renal progenitors in a step-wise fashion. We incorporated a priming step in our differentiation protocol in order to generate a monolayer and prime the cells for mesoderm differentiation. Microarray analysis at each step of differentiation showed that ESCs were differentiating towards the kidney lineage. At the end of the 21-day protocol, gene expression and functional studies demonstrated that our protocol yields a mixture of renal progenitors, including nephron progenitors, ureteric bud cells, and stromal progenitors. When aggregated and cultured on an air-liquid interface, the renal progenitors formed kidney organoids containing LTL+/E-cadherin+ proximal tubule cells and cytokeratin+ ureteric bud-derived tubules, and all the tubules were supported by extracellular matrix proteins produced by the cells themselves. A kidney organoid made up of primarily tubules is useful for drug-induced toxicity studies since many nephrotoxic drugs target tubules, in particular proximal tubules. The inability to fully mature organoids in vitro is an ongoing issue in the field, and, at best, researchers are only able to generate human kidney organoids in vitro that are equivalent to second trimester fetal kidneys 16 . We cultured our mouse organoids for up to 3 weeks of in vitro, mirroring the full mouse gestation time, and found that they did not fully mature to express organic anionic transporter 1 and 3 (Oat1, Oat3) and organic cation transporter 2 (Oct2), which are drug transporters commonly found on mature proximal tubule cells. Since mirroring gestation time was not sufficient to yield mature adult organoids, it suggests that developmental cue(s) may be missing from our culture.
Lastly, we investigated whether we could improve differentiation at the genetic level by over-expressing Pax2, Eya1, Six1, Hoxa11, Hoxc11, and Hoxd11 transgenes during ESC differentiation. With our system we found that over-expression of Pax2, Eya1, Six1, Hoxa11, Hoxc11, and Hoxd11 inhibited differentiation, indicating that the precise regulation of gene expression is critical for proper development. Having a robust differentiation protocol will allow researchers to study kidney differentiation and organogenesis in vitro, and therefore allows researchers to gain a better understanding of these complex processes. This understanding will ultimately help develop tools for human kidney disease diagnosis, management, and treatment.

Directed differentiation of mouse ESCs to renal progenitors
To recapitulate kidney development in vitro, we developed a stepwise protocol to direct the differentiation of mouse ESCs to mesoderm, intermediate mesoderm, and renal progenitors. First, we empirically established the optimal conditions, including growth factor combinations, the substrate on which cells are cultured and differentiated, and the seeding cell density for mesoderm induction. Bmp4 and activin A are strong inducers of mesoderm cells, and activin A and Fgf2 have been shown to upregulate brachyury (T) expression 3 . Different concentrations, combinations, and culture durations of Bmp4 (20-30 ng/ml), activin A (10-30 ng/ml), and Fgf2 (10-50 ng/ml) were tested and a combination of 30 ng/ml Bmp4, 10 ng/ml activin A, and 12 ng/ml Fgf2 for 2 days was found optimal for mesoderm induction (Fig. 1).
The extracellular matrix plays a critical role in differentiation, and thus gelatin and Geltrex, which are commonly used in cell culture and differentiation, were tested. Gelatin is composed of mainly collagen, while Geltrex is a mixture of extracellular matrix proteins. We plated mouse ESCs at 50,000 cells/cm 2 and differentiated them on gelatin-coated and Geltrex-coated tissue culture plates, and found that ESCs differentiated to T+ mesoderm more readily on gelatin-coated plates than Geltrexcoated plates (56.7 ± 6.4% vs. 22.3 ± 11.4% T+ cells, p < 0.05; Fig. 1).
Next, we explored whether adding a priming step in our differentiation protocol would improve differentiation efficiency. We hypothesized that the addition of a priming step consisting of Fgf2 and activin A to first prime the naive ESCs to epiblast stem cells (EpiSCs) would allow the cells to differentiate more readily to T+ mesoderm. In a study by Sudheer et al. 17 , they showed that culturing mouse ESCs in the presence of Fgf2 and activin A for 2 days was sufficient to convert ESCs to EpiSCs. Therefore, we plated mouse ESCs at 50,000 cells/cm 2 and differentiated mouse ESCs to T+ mesoderm with and without a 2-day priming step. We found that the percent T+ cells in prime and no prime conditions were not statistically different from each other (prime: 49.8 ± 12.2% versus no prime: 56.7 ± 6.4%; Fig. 1 (prime: 661/1366 T+ cells versus no prime: 175/303 T+ cells). Finally, the impact of seeding cell density on mesoderm induction was assessed. Cell-cell contact is important for cell-cell communication and therefore affects cellular differentiation. Too few or too many cells could affect differentiation. We seeded ESCs at three cell densities at the start of differentiation (12,500, 25,000 or 50,000 cells/cm 2 ) and found that ESCs seeded at 25,000 cells/ cm 2 yielded the most T+ cells after mesoderm induction (Fig. 1). When cells were seeded at 12,500, 25,000 and 50,000 cells/cm 2 , it yielded 37.3 ± 13.8%, 82.1 ± 17.0% and 49.8 ± 12.2% T+ cells, respectively. T induction was statistically higher (p < 0.05) when ESCs were seeded at 25,000 cells/cm 2 compared to 12,500 or 50,000 cells/cm 2 . These optimized initial conditions were then used to establish the remaining differentiation protocol.
Using quantitative polymerase chain reaction (PCR), we showed that our differentiation protocol yields a mix of renal progenitors, including metanephric mesenchyme (Osr1, Sall1), nephric duct (cRet), endothelial progenitors (CD34), and stromal progenitors (Foxd1, Pbx1) (n = 3; Fig. 2c). We found that the expression of these markers was absent/low in undifferentiated ESC, but increased as the ESCs differentiated to mesoderm, intermediate mesoderm, and progenitors. After differentiation, the progenitor cells expressed their respective markers at levels that were similar to E13.5 mouse embryonic kidney cells. Interestingly, at the end of the 21-day differentiation protocol, some cells expressed Cadherin 6 and Jag1, which indicated that some of the metanephric mesenchyme had differentiated further to renal vesicles. Finally, immunocytochemistry was used to reveal that pluripotencyassociated marker Oct4 was absent in ESC-derived progenitors, and quantitative PCR was used to show that pluripotencyassociated markers Esrrb, Lefty2, and Lin28a were significantly downregulated in progenitors compared to undifferentiated ESCs ( Supplementary Fig. 3). This indicated that the differentiated renal progenitors are not pluripotent.  4a). This shows that ESC-derived progenitors are more transcriptionally similar to embryonic kidneys than undifferentiated ESC, mesoderm, and intermediate mesoderm. Next, we used limma to determine significant differentially expressed genes between the different stages of differentiation, and we fed these lists into overrepresentation analysis to determine the enriched gene ontologies (GOs) associated with each stage of differentiation. Binary comparison of undifferentiated mouse ESCs to mesoderm cells revealed that 1941 genes were differentially expressed between them. Of the 1941 genes, 880 genes were significantly upregulated (false discovery rate (FDR)-corrected p value < 0.05) in mesoderm cells compared to undifferentiated ESC. Overrepresentation analysis of these genes revealed that GOs enriched in mesoderm cells included cell differentiation, development and morphogenesis, organ development and morphogenesis, and anterior/posterior pattern formation (Fig. 3a). These are all functions involved in gastrulation and the formation of mesoderm.
As the mesoderm cells differentiated to intermediate mesoderm cells, 1916 genes were differentially expressed with 881 genes significantly upregulated (FDR-corrected p value < 0.05) in intermediate mesoderm cells compared to mesoderm cells. Overrepresented GOs enriched in intermediate mesoderm cells include metanephros development, epithelium development, ureteric bud development, and vasculature development (Fig. 3b). This supports the notion that our differentiation protocol yields nephron progenitors, ureteric bud cells, and endothelial progenitors.
Finally, the differentiation of intermediate mesoderm cells to renal progenitors resulted in 2842 differentially expressed genes, of which 958 genes were significantly upregulated (FDR-corrected p value < 0.05) in progenitors compared to intermediate mesoderm cells. The over-represented GOs enriched in progenitors include kidney development, regulation of cell adhesion, and regulation of cell-matrix adhesion (Fig. 3c). The sustained enrichment of kidney developmental genes indicated that the progenitor induction media promoted the continued differentiation of renal progenitors.
Binary comparison of ESC-derived renal progenitors and E12.5/ E13.5 embryonic kidney would reveal genes differentially expressed between the samples and indicate what genes are lacking in each sample with respect to the other. Seven thousand and eight genes were differentially expressed between ESCderived renal progenitors and embryonic kidneys. Of the 7808 genes, 3201 genes were significantly upregulated (FDR-corrected p value < 0.05) in ESC-derived progenitors compared to embryonic kidneys, and 4607 genes were significantly upregulated in embryonic kidneys compared to progenitors. Over-represented GOs enriched in ESC-derived progenitors include extracellular matrix organization, regulation of cell migration, and cellular response to stress, hormone, organic substance, and cytokine ( Fig.  4a). On the other hand, GOs over-represented in embryonic kidneys include kidney development, nucleotide biosynthesis, and DNA replication, repair, and recombination ( Fig. 4b). Importantly, focusing on 190 metanephric mesenchyme-associated genes and 298 ureteric bud-associated genes, the gene expression profiles of renal progenitors were more similar to E12.5/13.5 mouse embryonic kidneys, then undifferentiated ESC, mesoderm cells, and intermediate mesoderm cells (Fig. 5).
Altogether, our microarray data suggest that mouse ESCs are differentiated towards the renal lineage in our differentiation protocol, and that the expression of kidney developmental genes increase with each step of differentiation, resulting in enrichment in kidney-related genes and GOs.
Renal progenitors give rise to kidney organoids when aggregated and cultured on an air-liquid interface Mouse embryonic kidneys have the intrinsic ability to self-organize into renal structures when dissociated into single cells, aggregated, and cultured as a cell aggregate on an air-liquid interface. These structured cell aggregates are also referred to as organoids. We dissociated and cultured E13.5 mouse embryonic kidney cells on a porous filter and showed that they give rise to N-cadherin+ proximal tubules and cytokeratin 8+ ureteric bud-derived structures ( Supplementary Fig. 5a). Similarly, when we dissociated our ESC-derived renal progenitors into single cells and cultured them on an air-liquid interface, they gave rise to the same structures as embryonic kidney cells ( Supplementary Fig. 5b   exogenous cells were added to the organoid culture so all the structures were derived from ESC-derived renal progenitors. A closer look at cross-sections of our organoids revealed that the tubules, indicated by E-cadherin and cytokeratin (pan), were surrounded and supported by extracellular matrix proteins, collagen IV, laminin, and fibronectin (Fig. 6). Since no extracellular matrix proteins or scaffolds were used during the aggregation and culture of the organoids, all the observed extracellular matrix proteins were produced by the renal progenitors themselves. In addition, we found Pbx1+ stromal cells in the space between tubules, which is where they are found naturally ( Supplementary  Fig. 6). The expression of E-cadherin, cytokeratin, and aquaporin 2 suggests that we have tubules from both metanephric mesenchyme and ureteric bud origins ( Supplementary Fig. 6). As a control, undifferentiated mouse ESCs were aggregated and cultured under the same conditions as kidney organoids to demonstrate that they do not give rise to the same tubule structures and extracellular matrix protein expression and organization as kidney organoids ( Supplementary Fig. 7). Secondary antibody controls confirmed that our secondary antibodies did not bind non-specifically to cells (Supplementary Fig. 8).
Next, we performed whole-mount staining on kidney organoids and cell aggregates made from mouse embryonic kidney, ESCderived renal progenitors, and undifferentiated ESCs (Fig. 7). Kidney organoids made from ESC-derived renal progenitors gave rise to lotus tetragonolobus lectin (LTL)+/E-cadherin+ mature Fig. 4 Network diagrams of selected gene ontology clusters enriched in ESC-derived renal progenitors compared to E12.5/13.5 mouse embryonic kidneys and vice versa. a Over-represented gene ontologies significantly enriched in progenitor cells compared to E12.5/13.5 embryonic kidneys include extracellular matrix organization and cellular response to stress, hormone, organic substance, and cytokine. b Over-represented gene ontologies significantly enriched in mouse embryonic kidneys compared to ESC-derived progenitors include kidney development, DNA replication, repair and recombination, and regulation of cell cycle. FDR-corrected p < 0.05. proximal tubules, but had few Wt1+ podocytes. On the other hand, kidney organoids made from E13.5 embryonic kidney cells had LTL+/E-cadherin immature proximal tubules, but many Wt1+ podocytes. This suggests that the organoids made from ESC-derived renal progenitors had more developed proximal tubules, but lacked podocytes compared to the organoids made from embryonic kidney cells. Our organoids are composed primarily of tubules because we added a Wnt agonist (CHIR99021) and Bmp7 to the organoid culture. Wnt and Bmp signaling play roles in patterning the kidney and determining the glomerulus-totubule ratio 25 . Similar to our results, work by Low et al. 26 and Garreta et al. 16 also showed that the addition of CHIR99021 results in a high tubule-to-glomerulus ratio. Lastly, as a control, cell aggregates from undifferentiated ESCs were made and they did   Fig. 4b). This indicates that the organoids are more similar to mouse embryonic kidneys than adult kidneys, but has commonality with the adult kidney for some of the expressed genes. Binary comparison of undifferentiated mouse ESCs to 8D organoids revealed that there were 369 differentially expressed genes (FDR-corrected p value < 0.05). Of the 369 genes, 195 genes were significantly upregulated in 8D organoids. GOs enriched in 8D organoids include renal system development, extracellular matrix organization, and blood vessel development, indicating that the ESCs differentiated to kidney cells (Fig. 8a). Between 14D organoids and mouse ESC, and 21D organoids and mouse ESC, there were 5017 and 5806 differentially expressed genes (FDRcorrected p value < 0.05), respectively. 14D and 21D organoids had similar over-represented GO enrichments as 8D organoids (data not shown). There were no significant differentially expressed genes between 8D and 14D organoids, and 14D and 21D organoids, suggesting that the organoids did not mature further in vitro with extended culture. This is not surprising considering that the only proven method to mature kidney organoids is to transplant them into mice for extended in vivo culture.
Comparing 8D organoids and adult mouse kidneys, there were 9244 differentially expressed genes (FDR-corrected p value < 0.05). Of the 9244 genes, 4843 genes were significantly upregulated in 8D organoids and 4401 genes were significantly upregulated in adult kidneys. Examples of over-represented GOs significantly enriched in 8D organoids include regulation of cell division, histone acetylation, histone methylation, metanephros development, and stem cell development and maintenance (Fig. 8b). Some of the over-represented GOs enriched in adult kidneys compared to 8D organoids include regulation of pH, and anion, cation, carboxylic acid, and amine transport, suggesting that the cells in the organoids lack these functions (Fig. 8c). Together, these comparisons suggest that the cells in the organoids are immature and still in the development stage, and they do not exhibit the same functions as adult kidneys. Between 14D organoids and adult kidney, and 21D organoid and adult kidney, there were 10,384 and 9810 differentially expressed genes (FDR-corrected p value < 0.05), respectively. The over-represented GOs enriched in 14D and 21D organoids compared to adult kidneys were similar to the GOs enriched in 8D organoids (data not shown). These results are expected considering there were no significant differentially expressed genes between 8D and 14D organoids, and between 14D and 21D organoids.
A characteristic of cellular maturity is the expression of drug transporter proteins on tubular cells. Cisplatin is an effective chemotherapy drug, but it is also nephrotoxic. Cisplatin-induced nephrotoxicity is mediated by Oct2 and Oat1 and Oat3. Oct2 mediates the uptake of cisplatin, while Oat1 and Oat3 mediate the uptake of mercapturic acid, which is a metabolite of cisplatin [27][28][29] . Our microarray data show that Oct2, Oat1, and Oat3 expression are significantly lower in 8D, 14D, and 21D organoids compared to adult mouse kidneys (Fig. 9a). This differential expression is also confirmed by quantitative PCR (Fig. 9b). Altogether, our gene expression data show that even with 3 weeks of culture, which mirrors mouse gestation time, there is still a large number of differentially expressed genes between organoids and adult kidneys, including drug transporters, indicating that our organoids are not at the same maturity level as adult kidneys. The inability to fully mature kidney organoids in vitro is an ongoing issue in the field.
Over-expression of Pax2, Eya1, Six1, Hoxa11, Hoxc11, and Hoxd11 during renal differentiation Pax2, Eya1, Six1, and Hox11 paralogs are transcription factors and co-activators involved in kidney development. We investigated whether the over-expression of Six1, Pax2, Eya1, Hoxa11, Hoxc11, and Hoxd11 in mouse ESCs would improve renal differentiation efficiency and possibly maturation. To do this, we cloned the genes into a doxycycline (DOX)-inducible piggyBac (PB) transposon system using gateway cloning. A DOX-inducible system gives us temporal control of transgene expression.
We introduced the transgenes into mouse ESCs using electroporation, and transgenic ESC clones were analyzed for the presence of the six transgenes (Supplementary Table I). Ten   Table II). The three most DOX-responsive clones (clones 11, 19, and 23) were selected for renal differentiation. The transgenic ESCs were differentiated to renal progenitors using our established protocol with DOX added at different time points during differentiation. DOX was added at days 6-11, 11-16 or 16-21 of differentiation, which coincided with the progenitor induction stage of differentiation. When DOX was added from days 6 to 11, Osr1, Jag1, and Pbx1 expression were significantly downregulated (p < 0.05) compared to the no DOX control group. On the other hand, Gata3 and Cited1 expression was significantly upregulated. When DOX was added from days 11 to 16, Osr1, cRet, Foxd1, Pbx1, Jag1, and Cadherin 6 expression were significantly downregulated compared to the no DOX group, while Gata3 continued to be significantly upregulated compared to the no DOX group. Lastly, when DOX was added from days 16 to 21, Osr1, Cited1, Wt1, Pbx1, Jag1, and Cadherin 6 were significantly downregulated compared to the no DOX group (Fig. 10). Altogether, our data suggest that over-expression of Six1, Pax2, Eya1, Hoxa11, Hoxc11, and Hoxd11 inhibits renal differentiation.

DISCUSSION
PSCs, which include ESCs and iPSCs, are capable of unlimited selfrenewal and differentiation to cells of all three germ layers. These properties make them a powerful tool for studying and modeling embryogenesis, organogenesis, and cellular differentiation and regeneration. One major challenge faced by researchers is having a reliable and reproducible method to direct the differentiation of PSCs to the target cell type(s). By mimicking embryonic kidney development, we developed a step-wise protocol to direct the differentiation of mouse ESCs to mesoderm, intermediate mesoderm, and renal progenitors that resemble embryonic kidney cells. Briefly, our differentiation protocol consists of culturing ESCs as a monolayer in four induction media: priming, mesoderm, intermediate mesoderm, and progenitor. Each induction medium contains growth factors and small molecules tailored to the specific stage of kidney development. Unlike other published mouse differentiation protocols that utilizes an embryoid body step, our monolayer differentiation strategy gives us better control of differentiation trajectory and also yields metanephric mesenchyme, ureteric bud cells, and stromal cells. In our protocol, mouse ESCs were differentiated on gelatincoated plates at a seeding cell density of 25,000 cells/cm 2 . Gelatin is made from hydrolyzed collagen, and coating culture plates with gelatin is known to mediate cell attachment. We also differentiated cells on Geltrex-coated plates because we hypothesized that differentiating cells on Geltrex would improve differentiation since Geltrex is made from a mixture of extracellular matrix proteins and may provide important proteins for attachment and differentiation that are missing from gelatin. Contrary to our expectations, differentiating cells on Geltrex did not improve mesoderm induction, and this could be because mouse ESCs are grown and maintained on gelatin-coated plates prior to differentiation. In addition, Geltrex is a ready-to-use solution containing  unknown growth factors, albeit at a low concentration, but they may be present at levels that interfere with mesoderm differentiation. We and others have highlighted the importance of cell seeding density in defining cell fate and differentiation efficiency [30][31][32][33] . Cell density influences cell-cell interaction and paracrine signaling between cells, and ultimately influences cellular differentiation. Here, we showed that mouse ESCs behaved differently when seeded at low, intermediate, and high cell density with the intermediate cell density yielding the most T+ cells. It is important to note that the optimal seeding cell density may vary between cell lines and should be optimized for each cell line.
Mouse ESCs are derived from the inner cell mass of E3.5 blastocysts and give rise to the visceral and parietal yolk sacs and the epiblast. EpiSCs give rise to cells of all three germ layers and can be found in E4.5 blastocysts. Although ESCs and EpiSCs are both pluripotent in nature, there are fundamental differences between the two cell types 34 . Morphologically, ESC colonies appear as tight, compact domes, while EpiSC colonies are flat. ESC maintenance is dependent on leukemia inhibitory factor (LIF), while EpiSCs require activin A and Fgf2. Prior to mesoderm induction, we included a priming step in our differentiation protocol to convert ESCs to EpiSCs because we hypothesized that priming would improve mesoderm differentiation efficiency 17,35 . Not only would a priming step prime ESCs for differentiation to mesoderm, but it would also flatten the cells into a flat monolayer, which allows for better control of differentiation trajectory. Indeed, we found that the differentiation protocol yields more T+ cells when a priming step was included.
We developed our differentiation protocol using the monolayer method because it has advantages over the embryoid body method. Monolayer culture allows for equal distribution of nutrients and signaling molecules among the cells. Embryoid bodies can vary in size and it is difficult to have consistently uniform embryoid bodies. We reasoned that size inconsistency and heterogeneity could lead to variations in the differentiated population. More importantly, we showed that ESCs differentiated as a monolayer can give rise to renal progenitors that could be expanded in culture and still retain the expression of progenitor markers Six2, Pax2, and Wt1 ( Supplementary Figs 9-11).
The step-wise structure of our differentiation protocol allowed us to examine genome-wide transcriptional changes at each step of differentiation. We found that the relative expression of kidney developmental genes increased with each step of differentiation, which resulted in enrichment for kidney-related GOs, including metanephros and ureteric bud development. This confirms that, unlike other mouse differentiation protocols, our protocol gives rise to both metanephric mesenchyme and ureteric bud progenitors. Binary comparison of ESC-derived progenitors and embryonic kidneys revealed that there were some fundamental differences between them. These differences lie in the level of kidney developmental gene expression, the extent of extracellular matrix organization and remodeling, and the cellular response to stimuli.
Stem cell organoids, or mini organs, are collections of stem/ progenitors cells that self-organize to form tissue-specific structures through processes that resemble in vivo tissue maturation, namely, cell sorting and spatially restricted lineage commitment 36 . Organoids were first made by dissociating and re-aggregating embryonic tissue, but using the same rationale, tissue-specific stem/progenitor cells differentiated from PSCs could also be coaxed to form organoids. Since PSCs can be genetically engineered to harbor disease-specific mutations or they can be made from patient cells, PSC-derived organoids are ideal models to study organogenesis, disease initiation and progression, and drug-induced nephrotoxicity in vitro.
When the ESC-derived renal progenitors were aggregated and cultured on an air-liquid interface, the progenitors gave rise to metanephric mesenchyme-derived proximal tubules and ureteric bud-derived cytokeratin+ tubules, and the tubules were supported by extracellular matrix proteins, collagen IV, laminin, and fibronectin, which were secreted by the cells themselves. The extracellular matrix not only provides structural support but it also serves as a reservoir for secreted growth factors. The presence of all three major extracellular matrix components is important for normal tissue development. No study has looked at the extracellular matrix composition of mouse organoids, and in terms of human organoids, groups have only surveyed for the presence of laminin. Here, we demonstrate that our mouse organoids contain collagen IV, laminin, and fibronectin.
We used microarray to study the transcriptional differences between ESC-derived organoids and undifferentiated ESCs, and between ESC-derived organoids and adult mouse kidneys. When compared to undifferentiated ESCs, genes associated with renal system development, blood vessel development, and extracellular matrix organization were significantly upregulated in the kidney organoids, as expected. We cultured our kidney organoids for up to 3 weeks in vitro and found that at the transcriptional level, there were no significant differences between 8D and 14D organoids, and 14D and 21D organoids. Even though our organoids contained proximal tubules that co-expressed LTL and E-cadherin, indicating maturity, we found that there were >9000 significant differentially expressed genes between organoids and adult kidneys, including drug transporter genes Oct2, Oat1, and Oat3. This suggests that our organoids, even with 3 weeks of culture that mirrored mouse gestation, are not at the same maturity level as adult kidneys. Our results are in line with other groups' results demonstrating that kidney organoids are difficult to mature in vitro and organoids often have to be transplanted into immunocompromised mice for long-term maturation [37][38][39][40] . With current technology, researchers are only able to generate human kidney organoids in vitro that are equivalent to second trimester fetal kidneys 16 .
It is speculated that it is difficult to mature organoids in vitro because culture conditions lack signaling molecules crucial for maturation and/or the length of in vitro culture has to mimic gestation time. Since mouse gestation is only 21 days, our mouse differentiation system allowed us to test whether mirroring gestation time enables kidney organoids to fully mature in vitro. In our system, we differentiated mouse ESCs to renal progenitors in 21 days, aggregated them to form kidney organoids and then cultured the kidney organoids for up to 3 weeks in vitro. Our culture time exceeds the length of mouse gestation, but our organoids did not fully mature. Therefore, it is more likely that our culture system is missing signaling molecules that are important for kidney maturation. Other groups have also shown that extended static in vitro culture of organoids does not improve organoid maturation and may be detrimental to the organoids 37,40 . Flow and shear stress have been implicated in cellular and tissue maturation, and considering that there is no blood flow in our organoids in vitro, this may be one of the limiting factors. Indeed, Homan et al. 41 showed that culturing kidney organoids under flow on microfluidic chips helps mature organoids.
Lastly, we described a system in which we can study the effects of gene over-expression on kidney differentiation. Although the effects of gene knockout or knockdown on kidney development are well studied, the effects of gene over-expression are not. We investigated whether over-expression of Pax2, Eya1, Six1, and the Hox11 paralogs could improve kidney differentiation. These genes form a gene regulatory network that ultimately activates and regulates Six2 expression and the progenitor cell population. Mutations in these genes result in kidney malformations or agenesis [5][6][7][8] . Given the importance of this gene expression network in kidney development, we hypothesized that overexpressing these genes would improve differentiation of ESCs to renal progenitors. Over-expressing six factors at once is not uncommon in the field of direct reprogramming or direct conversion of one cell type to another. For instance, Hendry et al. 42 over-expressed six factors at once (SIX1, SIX2, OSR1, EYA1, HOXA11, and SNAI2) to convert adult proximal tubule cells to nephron progenitors. In another example, Kaminski et al. 43 overexpressed four factors (Emx2, Hnf1b, Hnf4a, and Pax8) at the same time to convert fibroblast to renal tubular epithelial cells. In the most famous example of all, Takahashi and Yamanaka 44 initially over-expressed 24 genes at once to convert mouse fibroblasts to iPSCs before they performed an elimination experiment to identify the four indispensable genes.
Using quantitative PCR as a measure of differentiation efficiency, we assessed whether over-expression of these six genes during progenitor induction would improve differentiation efficiency. Contrary to our hypothesis, over-expression of Pax2, Eya1, Six1, and Hox11 paralogs inhibited differentiation as indicated by the marked reduction in the expression of progenitor markers, and since this did not yield renal progenitors, additional analysis such as cell aggregation and morphological assessments were not performed. We observed a marked reduction in progenitor marker expression because the relative expression levels and the timing of the six transgenes over-expression may be inappropriate for proper deployment of the kidney-specific gene expression network and/or the high doses of the six transgenes could be detrimental to the cell. In our system, it is difficult to control the relative expression levels of transgenes since we made the transgenic mouse ESC lines using the PB transposon system that inserts inverse terminal repeat-flanked transgenes into TTAA DNA sequences and so we have no control over the number of copies of transgenes inserted into the genome. Moreover, our system does not allow for sequential expression of the transgenes since all the transgenes are under the control of a DOX-inducible promoter. The next steps would be to investigate whether sequential expression of the six genes would improve differentiation.
In summary, we described in depth a strategy to differentiate mouse ESCs to renal progenitors and kidney organoids, and how this system can be modified to study the effects of gene overexpression on kidney differentiation. Our results highlight the importance of monolayer culture, seeding cell density, extracellular matrix, growth factor, and gene dose on kidney differentiation. In order to fully harness the potential of kidney organoids in modeling advanced diseases and nephrotoxicity, future studies need to focus on developing strategies to mature and pattern organoids to resemble adult kidneys.

Ethics statement
All experiments were approved by and performed at the Lunenfeld Tanenbaum Research Institute at Mount Sinai Hospital in Toronto, Canada. Only mouse cell lines were used in the experiments. No animals or human subjects/tissues were used.

Directed differentiation of mouse ESCs to renal progenitors
Prior to differentiation, mouse ESCs were cultured until they were 70-80% confluent. Confluent ESCs were dissociated into single cells with TrypLE Express (Invitrogen) and passed through a 70 µm cell strainer to obtain single cells. Cells were centrifuged at 200 × g, 10°C for 5 min. The supernatant was discarded and the cell pellet was resuspended in Priming Media (DMEM/F12 containing 1× N2 supplement, 1× B27 supplement, 0.1 mM NEAA, 2 mM GlutaMAX, 50 U/ml penicillin, 50 µg/ml streptomycin, 12 ng/ml Fgf2, and 20 ng/ml activin A). Cells were plated onto 0.1% gelatin-coated plates at a seeding cell density of 25,000 cells/cm 2 . Cells were cultured in priming media for 2 days before the cells were washed once with phosphate-buffered saline (PBS) and cultured in mesoderm induction media (DMEM/F12 containing 4% knockout serum replacement (KSR), 0.1 mM NEAA, 2 mM GlutaMAX, 50 U/ml penicillin, 50 µg/ml streptomycin, 1 mM sodium pyruvate, 0.015% w/v sodium bicarbonate, 0.1 mM 2-mercaptoethanol, 30 ng/ml Bmp4, 10 ng/ml activin A, and 12 ng/ ml Fgf2) for 2 days. After mesoderm induction, cells were washed once with PBS and cultured in an intermediate mesoderm induction media modified from Sambi et al. 24 (DMEM/F12 containing 4% KSR, 50 U/ml penicillin, 50 µg/ml streptomycin, 10 µM Y27632, 100 nM retinoic acid, and 30 ng/ml activin A) for 2 days 24  Quantitative PCR data was acquired and analyzed using the Bio-Rad CFX system. Expression data is displayed as the average of 3-5 replicates with standard deviation. Statistical significance between sample groups was calculated using unpaired t test (with Welch's correction) on GraphPad Prism 8 and p values < 0.05 (two tailed) were considered significant. Microarray RNA was isolated using the RNeasy Mini Kit (Qiagen) and RNA quality was assessed using the Agilent Bioanalyzer. Only RNA with a RNA integrity number of >7.0 were used for microarray. Samples were hybridized to Affymetrix Mouse Gene 2.0 ST array chips by a trained personnel at The Center for Applied Genomics (Toronto, Canada). Raw CEL files were read and normalized in R using the oligo and limma Bioconductor packages [46][47][48] . PCA plots were generated using ggplot2 49 . The limma package in R Bioconductor was used to determine significant differentially expressed genes for each comparison (FDR-corrected p value < 0.05). Gene set enrichment analysis (GSEA) was used for binary comparisons with <500 significant differentially expressed genes. GSEA was conducted using the R Bioconductor topGO package using the EMAPA database of mouse developmental anatomy ontology (http://www.obofoundry.org/ontology/ emapa.html) 50 . To narrow down the ontologies related to adult kidneys, we did a binary comparison between ESC and adult mouse kidneys to create a subset of ontologies highly enriched in their respective tissues. The subset of ontologies enriched to adult mouse kidneys was applied to experimental samples using topGO to determine if genes contributing to the ontologies are significantly enriched. Statistical significance is defined as having a FDR-corrected p value < 0.05.
Over-representation analysis was used for binary comparisons with >500 significant differentially expressed genes. For over-representation analysis, the gene lists were entered into BiNGO on Cytoscape to conduct over-representation analysis for GO Biological Processes (Benjamin and Hochberg FDR-corrected p value < 0.05, ontologies downloaded from www.geneontology.com on August 2018) 51,52 . The output was then exported and replotted using the Enrichment Map plugin for Cytoscape with the Node Cutoff set as 0.0001 and Edge Cutoff set as 0.7 53 . AutoAnnotate and WordCloud plugins were used to generate clusters of terms using Community Cluster (GLay) cluster algorithm and the label algorithm using four of the biggest keywords, respectively 54,55 . Only selected GOs are shown in figures.

Immunocytochemistry
The medium was removed and cells were washed twice in PBS. Cells were then fixed in 10% neutral-buffered formalin (NBF) for 20 min at room temperature (RT). Fixed cells were washed with PBS containing 0.1% Triton X-100, and incubated with blocking buffer (PBS containing 0.1% Triton X-100 and 10% FBS) for 30 min at RT. After blocking, primary antibodies (diluted 1:200) were added and incubated with the cells overnight at 4°C. The next day, cells were washed 4× for 10 min with PBS containing 0.1% Triton X-100 prior to incubating with secondary antibodies (diluted 1:500) for 30 min in the dark and at RT. Finally, cells were washed 5 × 10 min with PBS containing 0.1% Triton X-100, stained with DAPI (4′,6-diamidino-2phenylindole), and stored in PBS containing 50% glycerol, in the dark at 4°C until imaging. Secondary antibody controls in which only the secondary antibodies were added were performed to confirm the specificity of the secondary antibody. The primary antibodies used were as follows: T (Santa Cruz, SC-17743), Six2 (ProteinTech, 11562-1-AP), Wt1 (Abcam, AB89901), Pax2 (Covance, PRB276P), and Oct4 (Santa Cruz, SC8628). The secondary antibodies used were as follows: anti-goat IgG 594 and anti-rabbit IgG Cy5.

Microscopy
All images were taken on the Quorum WaveFX Spinning Disc Confocal System equipped with CSU X1 Confocal Scanner Unit (Yokogawa), ImagEM EM-CDD Camera (Hamamatsu), and a DMI6000 B Fully Automated Inverted Research Microscope (Leica). Images were acquired and processed on the Volocity Software Version 6.3.0 (Improvision/PerkinElmer).

Generating transgenic mouse ESC lines
ESCs were cultured under standard conditions and grown until 70-80% confluent. Confluent ESCs were collected, dissociated into single cells, and electroporated using the Neon Electroporation System (Invitrogen). Cells were electroporated under the following parameters: 1400 mV, 10 ms, and three pulses. Approximately 1 µg of PB transposase, rtTA, Pax2, Eya1, Six1, Hoxa11, Hoxc11, and Hoxd11 PB vectors were transfected into ESCs. Transfected ESCs were cultured under standard mouse ESC conditions. One day after transfection, puromycin (1 µg/ml) was added to culture for 24 h to select for cells containing the transgenes. The cells that survived puromycin selection were allowed to grow and form ESC colonies.

Differentiating transgenic mouse ESC lines to renal progenitors
Transgenic mouse ESC lines (clones 11, 19, and 23) were differentiated to renal progenitors following our differentiation protocol. DOX (1.5 µg/ml) was added days 6-11, 11-16, or 16-21 of differentiation, which coincided with the progenitor induction stage of differentiation. Differentiation with no DOX was also performed as a control. RNA was collected at the end of differentiation, cDNA was synthesized, and progenitor marker expression was quantified using quantitative PCR.

Organoid culture
At the end of the 21-day mouse kidney differentiation protocol, renal progenitors were dissociated into single cells and centrifuged at 200 × g at 10°C for 5 min. The supernatant was discarded and the cell pellet was resuspended in 15% FBS Complete Media (DMEM supplemented with 15% FBS, 50 U/ml penicillin, 50 µg/ml streptomycin, 2 mM GlutaMAX, 0.1 mM NEAA, 1 mM sodium pyruvate, and 0.1 mM 2-mercaptoethanol) at a concentration of 35,000-50,000 cells/µl. A 10 µl droplet of cells was transferred onto a porous membrane filter (Whatman, #110407) to form a cell aggregate. Cell aggregates were cultured overnight in 15% FBS Complete Media containing with 10 µM Y27632. The next day, cell aggregates were then cultured for 24 h in 15% FBS Complete Media containing 3 µM CHIR99021 and 10 ng/ml Bmp7. Finally, cell aggregates were cultured in 15% FBS Complete Media alone for 6 days (8D organoids), 12 days (14D organoids), or 19 days (21D organoids). Unless stated otherwise, 8D organoids were used in experiments.