ABSTRACT
Notch signaling promotes maturation of nephron epithelia, but its proposed contribution to nephron segmentation into proximal and distal domains has been called into doubt. We leveraged single cell and bulk RNA-seq, quantitative immunofluorescent lineage/fate tracing, and genetically modified human induced pluripotent stem cells (iPSCs) to revisit this question in developing mouse kidneys and human kidney organoids. We confirmed that Notch signaling is needed for maturation of all nephron lineages, and thus mature lineage markers fail to detect a fate bias. By contrast, early markers identified a distal fate bias in cells lacking Notch2, and a concomitant increase in early proximal and podocyte fates in cells expressing hyperactive Notch1 was observed. Orthogonal support for a conserved role for Notch signaling in the distal/proximal axis segmentation is provided by the demonstration that nicastrin (NCSTN)-deficient human iPSC-derived organoids differentiate into TFA2B+ distal tubule and CDH1+ connecting segment progenitors, but not into HNF4A+ or LTL+ proximal progenitors.
INTRODUCTION
The evolutionarily conserved Notch pathway translates an inter-cellular interaction into intra-cellular transcriptional outputs that control cell fate, proliferation, differentiation and apoptosis in a context-specific manner (Artavanis-Tsakonas et al., 1999; Bray, 2006; Kopan and Ilagan, 2009; Kovall et al., 2017). Mammals possess four Notch receptors (N1 to N4) and five delta (Dll)/jagged (Jag) ligands; all produce a signal when a bound ligand, presented and endocytosed by a neighboring cell, applies force that unfolds the Notch juxtamembrane negative-regulatory region, enabling cleavage by the metalloprotease ADAM10. The truncated, cell membrane-bound polypeptide is then cleaved by the γ-secretase complex that frees the Notch intracellular domain (NICD), which translocates into the nucleus (Kopan and Ilagan, 2009; Kovall et al., 2017). NICD associates with the DNA-binding protein CSL (CBF1/Suppressor of Hairless/LAG-1, also known as RBPJ in vertebrates) and recruits the co-activator mastermind to activate Notch target gene expression (Bray, 2016; Gordon et al., 2008; Kovall et al., 2017). Notably, each Notch receptor can be used only once, with NICD degrading after association with the transcription machinery (Kuang et al., 2020).
Mammalian kidneys develop from a Six2-expressing progenitor population induced by the ureteric bud (UB) to transition from self-renewal to differentiation via mesenchymal-to-epithelial transition (MET), forming a multitude of nephrons (Costantini and Kopan, 2010). Although there is broad agreement that the Notch pathway plays an important role in nephron formation (nephrogenesis), some controversy exists as to the specifics of its contributions to this process in mammals. After MET, cells form a pre-tubular aggregate (PTA) and start proliferating, forming a renal vesicle (RV) that contorts into the S-shaped body (SSB), where it is thought future nephron linages are already set (Lindstrom et al., 2018, 2021). Early reports using γ-secretase inhibitors (Cheng et al., 2003) or the hypomorphic PSEN1 transgene in γ-secretase null mutants (Wang et al., 2003) suggested that Notch signals are required in a temporal window that starts after MET and ends after the formation of the SSB, noting the paucity of nephron epithelia in these kidneys. Follow-up studies established that Notch2 provided the ‘lion's share’ of the signal (Cheng et al., 2007; Liu et al., 2013; Surendran et al., 2010a). Some Wnt4-positive RV formed in these animals, but the transition to SSB failed and development was arrested. Cdh1 expression in Lotus tetragonolobus lectin (LTL), Wt1 and Krt8 triple-negative epithelia resembling an incomplete SSB was used to argue that the few surviving Notch2-deficient epithelial cells were distal in character. The reason for the limited epithelial endowment in Notch2 nulls was not explored beyond the reduced proliferation in Jag1+, Notch2–/– cells relative to wild-type cells. The assertion that Notch activity biased a distal default program towards a proximal tubule (PT) fate was further supported by activating a stabilized (degron-deleted) version of the Notch1 intracellular domain (N1ICD), the active form of the receptor, which resulted in the en masse differentiation of nephron progenitor cells (NPCs) into multiple LTL- or Wt1-positive proximal epithelial clusters lacking distal tubule or loop of Henlé (LOH) markers [based on explant cultures (Cheng et al., 2007) or RNA in in situ hybridization at embryonic day 15 (E15) (Boyle et al., 2011)].
Independent analysis of an activated, degron-containing N2ICD transgene (Fujimura et al., 2010) led to a somewhat different conclusion. Although Fujimura et al. also observed abundance of proximal tubules and aborted branching that was consistent with precocious NPC differentiation, they also report a greater degree of differentiation with abnormal tubular and glomerular cysts. The glomerular disruption was worse than in N1ICDΔPEST overexpression (Sweetwyne et al., 2015; Waters et al., 2008). Notably, although distal markers were not examined, they concluded that Notch signals were required for epithelial maturation, not fate selection. They suggested that because Notch1 and Notch2 promote different outcomes, they may play different roles. The similarity of the NICD overexpression phenotypes to the precocious differentiation observed in Six2 null mutants led them to propose that Notch target genes may contribute to epithelization both indirectly (by repressing the NPC maintenance factor Six2) and directly (by promoting maturation).
The idea that Notch1 and Notch2 play distinct developmental roles due to divergent residues in their respective intracellular domains was tested in vivo by swapping the N1ICD with the N2ICD and vice versa (Liu et al., 2013). Pax3Cre-mediated deletion of Notch2f/f could not be rescued by substituting N2ICD into Notch1. By contrast, mice without N2ICD (now having four copies of N1ICD, two replacing N2ICD) displayed normal kidney development. Because each Notch receptor is consumed as it generates a signal and cannot be reused, and because Notch2/Jag1 interactions produced more NICD polypeptides than Notch1/Jag1 interactions, it was concluded that signal strength, defined as the sum of NICD released from all ligand-bound Notch receptors on the cell surface, was a far more important determinant of Notch signaling outcomes in the kidney than NICD composition (Liu et al., 2015, 2013; Ong et al., 2006), providing a different interpretation for the differences between N1ICDΔPEST (strong) and N2ICD (weaker) overexpression.
Several mechanisms have been found to modulate Notch1 versus Notch2 signal strength, including receptor glycosylation (Haltiwanger, 2008; Kakuda and Haltiwanger, 2017), which generates the preferred Notch2/Jag1 and Notch1/Dll1 signaling pairs, explaining the more profound impact of deleting Jag1 than Dll1 (Liu et al., 2013). More recently, ligand-dependent signal dynamics were shown to be another key determinant of signaling outcomes (Nandagopal et al., 2018), with Dll1, and thus Notch1, signaling in a pulsatile pattern, whereas Jag1, and thus Notch2, elicit a sustained signal. Collectively, these observations are consistent with the idea that the different outcomes described above reflect differences in the relative strength of Notch signals, be it endogenous or overexpressed.
Whereas support for the role of Notch signaling in epithelial maturation has strengthened, Notch promotion of proximal fates over distal ones has been called into doubt (Chung et al., 2017; Fujimura et al., 2010). Chung et al. concurred that MET can be completed in Notch1;Notch2 or Rbpj nulls, but these kidneys had even fewer epithelia (Chung et al., 2016; Surendran et al., 2010a) than kidneys with Notch2 null NPCs (Cheng et al., 2007). Moreover, because the commonly used Six2+/Cre line is mosaic, allowing some NPCs to ‘escape’ deletion altogether or delay it sufficiently, formation of epithelia in the Surendran study could reflect such escapers. Another confounder of earlier findings is the use of Pax3Cre, which deletes Notch2 in both epithelia and mesenchyme. To address some of these issues, investigators examined differentiation markers for podocytes (Nphs2), proximal tubules (Slc34a1), distal tubules (Slc12a3) and LOH (Slc12a1) in mice by deleting both Notch1 and Notch2 with Wnt4+/Cre, and reported that no recognizable epithelial lineage formed from the SSB (Chung et al., 2017). Moreover, when N1ICDΔPEST was activated in nascent epithelia by Wnt4+/Cre, a cell fate bias was not observed based on these markers. The authors concluded that Notch signals have no role in segmenting the proximo-distal axis in nephrons, with a key role for Notch in transitioning from an epithelial vesicle to a nephron. They did not consider the possibility that, as Notch is required for maturation, the absence of late markers in loss-of-function models may not be informative.
The explosion of scRNA-seq studies led to the appreciation that first-generation markers, including the ones used by Chung et al. (2017), describe relatively mature cell populations. A more granular view of nephrogenesis emerged, with markers identifying intermediate steps (Naganuma et al., 2021). Using our own scRNA-seq data, we positioned Notch receptor-ligand pairs in a temporal context along the differentiation axis, and revisited the original Notch2 loss-of-function model with new tools. We demonstrate herein that deletion of Notch2 with Six2+/Cre, which deletes target genes in the NPCs and their epithelial descendants, sufficed to arrest epithelial development at an immature state, indistinguishable morphologically from our earlier report (Cheng et al., 2007). Notch2 antibody staining and lineage analysis confirmed that the epithelial cells that formed were descendants of Notch2-deficient cells. Most Notch2-deficient, GFP-positive descendants of Six2+/Cre; Notch2f/f; ROSA+/EYFP NPCs expressed Cdh1 and Tfap2b, an early marker for committing distal tubules/LOH (Marneros, 2020; Naganuma et al., 2021). A small minority of GFP+ cells were positive for LTL and Hnf4a, a committing proximal tubule marker (Marable et al., 2020), but not for Cdh1. The inverse bias favoring the proximal fate was seen in Six2+/Cre; N1ICDΔPEST kidneys, with abundant expression of Hnf4a+ cells and Wt1+ cells, but a lack of Tfap2b+ cells. A handful of epithelial cells expressed all lineage markers (LTL, Hnf4a, Wt1 and Tfap2b). Similarly, Six2+/Cre; Rbpjf/f kidneys had far more Tfap2b+ cells than LTL/Hnf4a+ cells. Delayed deletion of Rbpjf/f with Wnt4+/Cre did not compromise nephron formation, marking the end of the Notch-dependent window as the time required for RBPJ protein to turn over after the gene is deleted, with a half-life post-deletion measured at 60 h in the epidermis (Turkoz et al., 2016). At the transcriptome level, RNA-seq analysis confirmed a shift in gene expression towards early PT (ePT), and away from NPCs, mature PT, DT or LOH in N1ICDΔPEST-expressing NPCs. Finally, human kidney organoids derived from nicastrin (NCSTN)-deficient human induced pluripotent stem cells (iPSCs) without γ-secretase activity differentiated into KRT8/18- and TFAP2B-positive cells, but not into HNF4A- or LTL-positive cells. Combined, these analyses confirm a role for Notch signaling in proximal/distal nephron segmentation downstream of a more global role in promoting epithelial growth and maturation in both mouse and human.
RESULTS
scRNA-seq places the Notch signaling pathway in early epithelia with a slight proximal distribution bias
As introduced above, the present understanding of Notch signals in nephrogenesis based on markers for mature cell fates assumes that Notch signaling did not contribute to the epithelial fate selection process. Single cell RNA sequencing (scRNA-seq) suggested that the choice of markers confounded the interpretation. Integration analysis of six cortically biased scRNA-seq datasets (pooled embryos from three genotypes with high transcriptional similarity, including E14 and P0 timepoints) using the Seurat 3.0 UMAP algorithm (Stuart et al., 2019) identified several super clusters, characterized as NPCs, stromal and ureteric bud (UB, Ret+) ‘continents’ (Jarmas et al., 2021). The NPCs were aligned in a continuum with (Cited1+, Six2+, Wt1+, Nr2f2+ and Pax8+) naïve cells at one pole, followed by a transitional (Cited1−, Six2+, Wt1+, Nr2f2+, Wnt4+, Pax8+ and Lhx1+) population (marked with red brackets), and segregating into epithelial renal vesicle and precursors of distal tubule (DT; Fgf8+, Sim1+, Tfap2b+ and Gata3+), proximal tubule (PT; Cdh6+, Hnf4a+ and Fut4+), and podocyte (Pod; Mafb+) at the opposite pole (Fig. 1). Importantly, within each emerging epithelial cluster, the UMAP algorithm arranged cells from the least mature to the more mature, with the most mature markers [Slc12A3 (DT), Slc34a1 (PT) and Nphs2 (POD)] absent from this cortically biased dataset (Fig. 1). Interrogating the expression of Notch pathway ligands (Jag1 and Dll1), receptors (Notch1 and Notch2) and targets (Nrarp, Hes5, Hnf1b, Hey1, Hey2 and Heyl) shows expression in the transition zone and a strong distribution bias towards the proximal lineage, with minimal overlap with the Tfap2b domain (a marker for distal fates, Fig. 1). This pattern rekindled our interest in clarifying the role of Notch in nephron segmentation.
scRNA places the Notch pathway at PT/DT fate branching. Single cell RNA integration analysis of pooled of E14 and P0 embryos from Six2-cre, Six2 KI and Six2-cre; Tsc1+/f mice (Jarmas et al., 2021). The top row (left) orients the reader to the mesenchymal-epithelial (HoxC10-Epcam) axis and marks the cell population by coloring marker genes within the NPC cluster. Cell cycle phase is also shown. The naïve NPC region (Cited1+) is enriched for G1. Red dotted lines indicate the transition zone (Wnt4+) from committed NPCs (NPCc) to DT (Epcam+, Gata3+) and ePT (Hnf4a+) cells. In all panels, the area above the dashed red line contains the NPCs (Cited1+ and Six2+). Gene expression patterns are set such that all expressing cells are in front of non-expressing cells. Below the lower red dashed line, cells are transitioning to epithelia. The middle row shows the early marker for epithelial DT (eDT), PT (ePT) and POD. The bottom row depicts expression of Notch pathway members. The Notch1/Dll1 and Notch2/Jag1 are preferred receptor/ligand pairs. Notch target genes may respond also to other stimuli. Nrarp is considered a universal Notch-responsive target.
scRNA places the Notch pathway at PT/DT fate branching. Single cell RNA integration analysis of pooled of E14 and P0 embryos from Six2-cre, Six2 KI and Six2-cre; Tsc1+/f mice (Jarmas et al., 2021). The top row (left) orients the reader to the mesenchymal-epithelial (HoxC10-Epcam) axis and marks the cell population by coloring marker genes within the NPC cluster. Cell cycle phase is also shown. The naïve NPC region (Cited1+) is enriched for G1. Red dotted lines indicate the transition zone (Wnt4+) from committed NPCs (NPCc) to DT (Epcam+, Gata3+) and ePT (Hnf4a+) cells. In all panels, the area above the dashed red line contains the NPCs (Cited1+ and Six2+). Gene expression patterns are set such that all expressing cells are in front of non-expressing cells. Below the lower red dashed line, cells are transitioning to epithelia. The middle row shows the early marker for epithelial DT (eDT), PT (ePT) and POD. The bottom row depicts expression of Notch pathway members. The Notch1/Dll1 and Notch2/Jag1 are preferred receptor/ligand pairs. Notch target genes may respond also to other stimuli. Nrarp is considered a universal Notch-responsive target.
Notch2 deletion only in NPCs is sufficient to ablate nephrons and compromise viability
After establishing our mouse colony at Cincinnati Children's Hospital Medical Center (CCHMC), survival analysis of 10 litters (73 pups) confirmed that although Six2+/Cre; Notch2f/f pups were born at the expected Mendelian ratios, all perished by postnatal day 2 (P<10–5; Fig. 2A), exactly as reported for the Pax3+/Cre; Notch2f/f pups. The complete penetrance of the phenotype argued against the presence of an unlinked lethal allele, and using multiple unrelated Notch2f/f dams with no surviving Six2+/Cre; Notch2f/f pups argued against a linked mutation at the Notch2 locus. To rule out the possibility that our Six2+/Cre breeding stock acquired a linked mutation segregating with Six2-Cre, we obtained a stud from Dr Joo-Soep Park at CCHMC but, again, all Six2+/Cre; Notch2f/f pups perished by P2. This result demonstrated that Notch2 deletion only in NPCs sufficed to reproduce the observations made by Chung et al. (2017), eliminating the concern that Notch2 loss in mesenchyme contributed to the reported phenotypes.
Survival, histology and immunostaining of mice with Notch2-deficient nephrons. (A) Survival analysis. Six2+/Cre Notch2f/f mice were born at normal mendelian ratios, but none survived to P2 (asterisk). (B) Hematoxylin and Eosin stains of Six2 mutant and wild-type P1 kidneys. The red circle indicates an area of PT cyst formation in the mutant. (C) Six2 mutant and wild-type kidneys from littermates euthanized at P0 were stained using GFP (lineage trace), and for Notch2, LTL (PT marker) and Krt8/18 (a UB and CD marker). Dashed white rectangles are enlarged in C′,C″. (C′, top) View of the marginal zone region from a Six2+/Cre Notch2f/f with GFP (green) activated by Cre, and Notch2 (red) detected only in surrounding GFP-negative cells. (C′, bottom) Same image as above, showing only the Notch2 stain. White dot identifies similar structures in both panels. (C″, top) View of the marginal zone region from Six2+/Cre Notch2+/f (wild type) with GFP (green) activated by Cre and Notch2 (red). (Bottom) Same image as above, showing only Notch2. White dots identify similar structures in both panels. Scale bar: 100 µm.
Survival, histology and immunostaining of mice with Notch2-deficient nephrons. (A) Survival analysis. Six2+/Cre Notch2f/f mice were born at normal mendelian ratios, but none survived to P2 (asterisk). (B) Hematoxylin and Eosin stains of Six2 mutant and wild-type P1 kidneys. The red circle indicates an area of PT cyst formation in the mutant. (C) Six2 mutant and wild-type kidneys from littermates euthanized at P0 were stained using GFP (lineage trace), and for Notch2, LTL (PT marker) and Krt8/18 (a UB and CD marker). Dashed white rectangles are enlarged in C′,C″. (C′, top) View of the marginal zone region from a Six2+/Cre Notch2f/f with GFP (green) activated by Cre, and Notch2 (red) detected only in surrounding GFP-negative cells. (C′, bottom) Same image as above, showing only the Notch2 stain. White dot identifies similar structures in both panels. (C″, top) View of the marginal zone region from Six2+/Cre Notch2+/f (wild type) with GFP (green) activated by Cre and Notch2 (red). (Bottom) Same image as above, showing only Notch2. White dots identify similar structures in both panels. Scale bar: 100 µm.
Hematoxylin and Eosin (H&E) staining of postnatal day (P) 1 Six2+/Cre; Notch2f/f and Six2+/Cre; Notch2+/f control littermates showed greatly diminished nephrogenesis in the mutants (Fig. 2B, note the few nephrons forming due to mosaicism) that was indistinguishable from the description of Cheng et al. (2007). Immunofluorescence performed on P0 kidneys from Six2+/Cre; Notch2f/f; Rosa+/EYFP and Six2+/Cre; Notch2+/f; Rosa+/EYFP controls detected GFP (lineage trace), Notch2, LTL (PT marker) and Krt 8/18 [UB and collecting duct (CD) marker] with the specified antibodies (see Materials and Methods). As reported earlier, some proximal tubules were identified in Notch2-deficient kidneys, but only a few cells were marked with GFP, suggesting that these epithelial cells arose from Notch2-expressing NPCs. By contrast, nearly all nephron epithelia and NPCs in control mice were positive for GFP (Fig. 2C). Notch2, which is abundant in controls (Fig. 2C″), was not detected in NPCs of GFP+ epithelia of Six2+/Cre; Notch2f/f; Rosa+/EYFP kidneys (Fig. 2C′). Finally, to ask whether the milder phenotype reported previously reflected inefficient Cre-mediated recombination, we generated Fgf20+/Cre; Notch2f/f pups. Fgf20 is expressed at low levels in NPCs and other epithelia, such as the skin appendages and mammary glands (Barak et al., 2012; Elo et al., 2017; Huh et al., 2020, 2013, 2015; Yang et al., 2018), and our previous analyses suggested it is less efficient than Six2+/Cre (Mukherjee et al., 2021; Volovelsky et al., 2018). Fgf20+/Cre; Notch2f/f mice were born at normal Mendelian ratios, survived with no physical impairments, and were used for breeding. H&E staining of Fgf20+/Cre; Notch2f/f mice at P0 identified cysts in the kidneys with significantly fewer nephrons than controls (Fig. 3A,B). The survival of Fgf20+/Cre; Notch2f/f mice, the low nephron numbers, and the appearance of cysts were all reminiscent of what we reported previously for the Six2+/Cre; Notch2f/f animals (Surendran et al., 2010a,b). This suggests that the deletion efficiency in Six2+/Cre mice may be sensitive to unidentified environmental factors.
Histology, nephron counts and renal function analysis in Fgf20+/Cre and Foxd1+/Cre Notch2-deleted kidneys. (A) Histological analysis of Notch2 f/f; Fgf20+/Cre, Notch2+/f; Fgf20+/Cre, Notch2f/f; Foxd1+/Cre and Notch2+/f; Foxd1+/Cre kidneys. (B) Nephron counts in these strains. Data are presented as sample group mean±s.d. (n=6 kidneys per group for Fgf20+/Cre;N2f/f, Fgf20+/Cre;N2+//f and N2f/f; n=5 kidneys per group for Foxd1+/Cre;N2f/f and Foxd1+/Cre). For the Fgf20 groups, statistical significance was evaluated by one-way ANOVA with Tukey's multiple comparison tests: ***P≤0.001. For the Foxd1 groups, lack of statistical significance was demonstrated by an unpaired two-tailed t-test, performed in GraphPad Prism 8. (C) Renal function in these strains after P90; n=3 for Fgf20+/Cre;N2f/f, Fgf20+/Cre;N2+/f, Foxd1+/Cre;N2f/f, Foxd1+/Cre and N2f/f. All multiple t-tests and raw data are presented in Table S1. All multiple t-test adjusted values were greater than 0.5. Scale bars: 100 µm.
Histology, nephron counts and renal function analysis in Fgf20+/Cre and Foxd1+/Cre Notch2-deleted kidneys. (A) Histological analysis of Notch2 f/f; Fgf20+/Cre, Notch2+/f; Fgf20+/Cre, Notch2f/f; Foxd1+/Cre and Notch2+/f; Foxd1+/Cre kidneys. (B) Nephron counts in these strains. Data are presented as sample group mean±s.d. (n=6 kidneys per group for Fgf20+/Cre;N2f/f, Fgf20+/Cre;N2+//f and N2f/f; n=5 kidneys per group for Foxd1+/Cre;N2f/f and Foxd1+/Cre). For the Fgf20 groups, statistical significance was evaluated by one-way ANOVA with Tukey's multiple comparison tests: ***P≤0.001. For the Foxd1 groups, lack of statistical significance was demonstrated by an unpaired two-tailed t-test, performed in GraphPad Prism 8. (C) Renal function in these strains after P90; n=3 for Fgf20+/Cre;N2f/f, Fgf20+/Cre;N2+/f, Foxd1+/Cre;N2f/f, Foxd1+/Cre and N2f/f. All multiple t-tests and raw data are presented in Table S1. All multiple t-test adjusted values were greater than 0.5. Scale bars: 100 µm.
To determine whether loss of Notch2 in the stroma or reduced nephron numbers had untoward effect on long-term renal health, we tested renal function in Foxd1+/Cre; Notch2f/f and Fgf20+/Cre; Notch2f/f mice aged between 60 and 90 days. Foxd1 is expressed in the stroma and later in podocytes (POD) (Boyle et al., 2014). Like Fgf20+/Cre; Notch2f/f mice, Foxd1+/Cre; Notch2f/f mice were born at normal Mendelian ratios, survived with no physical impairments, and were used for breeding. Foxd1+/Cre; Notch2f/f kidneys and nephron numbers appeared indistinguishable from controls (Fig. 3B), as reported earlier (Boyle et al., 2014). To measure renal function, 0.5 ml of blood were collected per animal (n=3 for Foxd1+/Cre; Notch2f/f; Foxd1+/Cre; Notch2+/f and Fgf20+/Cre; Notch2f/f; Fgf20+/Cre; Notch2+/f; Notch2f/f) and sent to a diagnostic laboratory (IDEXX) for analysis. No significant differences between mutant and control littermates were detected, indicating normal renal function (Fig. 3C, Table S1). Thus, the threshold for formation of a kidney with sufficient function to support viability in mice can be satisfied with a delay in Notch2 deletion timing.
Quantifying fate choices of Notch2-deficient cells confirms preferential loss of proximal cell fates
To characterize and quantify the cell types generated by Notch2-deficient NPCs, mutant and control kidneys were stained with antibodies to Tfap2b, an early distal tubule (eDT) marker, to GFP, a lineage tracer, and to the lectin LTL, a PT marker. The tissue was imaged on a confocal microscope, and analyzed in Imaris (Fig. 4). Independently, we stained adjacent sections with antibodies to GFP, LTL and Hnf4a [an early proximal tubule (ePT) marker] to confirm that we did not miss any Hnf4a+ populations in Notch2-deficient kidneys (Fig. S2). A control littermate (Fig. 4A) demonstrates that all nephron epithelia were derived from NPCs in which GFP was activated following Cre-mediated recombination (note the transition from LTL to Tfap2b in the LOH, Fig. 4A′). The stained Notch2-deficient tissue from three individuals was initially divided into ∼20 arbitrary sectors radiating from the papilla, totaling 71 sectors in three mutant kidneys (Fig. 4B, Table S2). Three observers (one blinded to the experimental goal) counted the sectors that contained LTL alone, LTL and GFP, Tfap2b alone, Tfap2b and GFP, or LTL and Tfap2b, counting the last even if only one LTL-positive cell was detected. The counts were averaged to determine standard deviation. The percentage of sectors that contained Tfap2b+ (DT, 78.4%) was significantly higher than the percentage that contained LTL+ (PT, 58.69%; Fig. 4C, P=0.013). Sectors containing Tfap2b+ and LTL+ cells were considered GFP positive if 50% or more of the cells in the cluster were GFP+. By that definition, a significantly higher percentage of DT sectors were GFP+ than PT sectors, 44.6% to 13.6%, respectively (Fig. 4C, P=0.007), suggesting that in the absence of Notch, surviving epithelia are distal in character. If the differences in outcome of Notch2 deletion were attributed to differential proliferation of DT cells, DT cells will outnumber PT cells but each sector with DT cells should have at least a few PT cells.
Notch signals are necessary to promote PT fate. (A) Control kidneys stained for GFP (lineage tracer), LTL (PT) and Tfap2b (DT). The outlined region from one kidney is shown in a′. (B) Similarly stained Six2+/Cre; Notch2f/f kidneys from three animals were divided into 20 or 21 arbitrary sectors (n=71 sectors). (C) The 71 sectors shown in A were scored by three independent observers for the presence or absence of any of the three markers and the scores averaged. (D,E) To obtain quantitative data, the kidneys were imaged in IMARIS, segmented by color (Db-d, top) and the area deemed positive for each color presented in pseudocolor. In Da, the total kidney area (pink, asterisk in Dd marks the section used for this determination) and all the color combination are presented. (E) The area positive for each fate marker in D was calculated by the software as a fraction of the total kidney area and plotted. Data are mean±s.d. Scale bars: 100 µm.
Notch signals are necessary to promote PT fate. (A) Control kidneys stained for GFP (lineage tracer), LTL (PT) and Tfap2b (DT). The outlined region from one kidney is shown in a′. (B) Similarly stained Six2+/Cre; Notch2f/f kidneys from three animals were divided into 20 or 21 arbitrary sectors (n=71 sectors). (C) The 71 sectors shown in A were scored by three independent observers for the presence or absence of any of the three markers and the scores averaged. (D,E) To obtain quantitative data, the kidneys were imaged in IMARIS, segmented by color (Db-d, top) and the area deemed positive for each color presented in pseudocolor. In Da, the total kidney area (pink, asterisk in Dd marks the section used for this determination) and all the color combination are presented. (E) The area positive for each fate marker in D was calculated by the software as a fraction of the total kidney area and plotted. Data are mean±s.d. Scale bars: 100 µm.
To provide a more accurate accounting of fate choices, we created a binary algorithm in Nikon Elements to measure the kidney area (Fig. 4Da), the area covered by PT cells (LTL+, Fig. 4Db), the area covered by DT cells (Tfap2b+, Fig. 4Dc) and the area covered by GFP (Fig. 4Dd). Using these masks, we calculated the area that was LTL+/GFP+ and Tfap2b+/GFP+ in µm2, and extracted its percentage from the total DT and PT areas (Fig. 4E). In control Notch2+/f littermate kidneys, 86.6% of LTL+ and 96.8% of Tfap2b+ cells were also GFP+ (Fig. 4E, Table S2). In Six2+/Cre; Notch2f/f-deficient kidneys, only 12.4% of LTL+ cells were GFP positive, in contrast to 58.5% of Tfap2b+ cells that were also GFP positive. Thus, Notch2-deficient cells are more than four times as likely to express the early DT marker than they are to express the PT marker. A similar choice bias was observed in Fgf20+/Cre; Notch2f/f kidneys, with only 3.4% of LTL+ cells being GFP+ versus 40.3% of Tfap2b+ cells expressing GFP. It should be noted that, although Six2+/Cre; Rbpjf/f kidneys contain very few epithelial cells, 42.9% of 71 sectors were TFAP2b+; LTL− but only 0.42% were LTL+; TFAP2b− (P=0.014; Fig. S3A,B, Table S2). Notably, and in contrast to Wnt4+/Cre; Notch1f/f; Notch2f/f (Chung et al., 2017), Wnt4+/Cre; Rbpjf/f kidneys were indistinguishable from Wnt4+/Cre; RBP+/f kidneys (Fig. S3C), suggesting that perdurance of RBPJ protein (Turkoz et al., 2016) may suffice to complete nephrogenesis. Wnt4+/Cre; Rbpjf/f mice did not survive (Fig. S3D), however, perhaps due to the requirement of Notch signaling in other Wnt4-expressing tissues (Caprioli et al., 2015), a possibility we did not pursue further. These analyses complement our earlier published conclusion, inferred based on the patterns of LTL and C1 (Cheng et al., 2007), that Notch2 expression guides cells towards a proximal fate before acting in all cells to promote proliferation/maturation.
The stabilized Notch1 intracellular domain expands the PT at the expense of NPCs and the DT
When N1ICDΔPEST was overexpressed in Six2-expressing progenitors, it drove epithelialization in the absence of Wnt4 and Wnt9b, producing mostly LTL+, Cdh1− and Wt1+ cells (Boyle et al., 2011; Cheng et al., 2007). To determine whether early DT cells are also present under these conditions, we crossed a Six2+/Cre sire with RosaNICD/NICD dams (Murtaugh et al., 2003) and analyzed nephrogenesis histologically at E15.5, as we did previously (Boyle et al., 2011; Cheng et al., 2007), using additional markers and obtaining bulk RNA-seq data. The kidneys, arrested after one or two branching events (Fig. S4), were isolated and processed for whole-mount immunofluorescence with antibodies against Wt1 (NPC and podocyte marker), Hnf4a (PT marker), Tfap2b (DT marker) and GFP (co-expressed with N1ICDΔPEST). We again noted large expansion of epithelia and paucity of Six2+ NPCs but could only detect a few cells that expressed Tfap2b (Movies 1 and 2). These were also positive for Wt1, LTL and Hnf4a (Movie 1). GFP+/LTL+/Hnf4a+ triple-positive cells dominated (Fig. 5A-C, Movies 1 and 2). Interestingly, Wt1+ cells did not stain for GFP, suggesting they may have lost NICD expression, or arose from cells in which NICD was not activated by the Six2+/Cre allele. The lineage tracing by Chung et al. (2017) supports the former hypothesis.
N1ICDΔPEST promotes differentiation and proximal development at the expense of distal cells. (A) Whole mount of a kidney anlage expressing N1ICDΔPEST. The Wt1+ cells (green) are not expressing GFP (red). (B) A frame from Movie 1. (C) A frame showing Hnf4a expression from Movie 2. (D) P0 kidneys with the genotypes noted were dissociated and GFP+ cells (Six2+ NPC) counted by flow cytometry. Data are presented as sample group mean±s.d. (n=9 control Six2+/Cre; N2+/f and n=4 experimental Six2+/Cre; N2f/f isolated from two litters). P=0.5561 (not significant; n.s.) determined by an unpaired two-tailed t-test performed in GraphPad Prism 8. (E) Bisque deconvolution assigned relative cell type enrichment in bulk RNA samples from three wild-type and two N1ICDΔPEST-expressing kidneys using the top 20 or 100 markers from scRNA dataset (Combes et al., 2019) as a standard. Scale bar: 100 µm.
N1ICDΔPEST promotes differentiation and proximal development at the expense of distal cells. (A) Whole mount of a kidney anlage expressing N1ICDΔPEST. The Wt1+ cells (green) are not expressing GFP (red). (B) A frame from Movie 1. (C) A frame showing Hnf4a expression from Movie 2. (D) P0 kidneys with the genotypes noted were dissociated and GFP+ cells (Six2+ NPC) counted by flow cytometry. Data are presented as sample group mean±s.d. (n=9 control Six2+/Cre; N2+/f and n=4 experimental Six2+/Cre; N2f/f isolated from two litters). P=0.5561 (not significant; n.s.) determined by an unpaired two-tailed t-test performed in GraphPad Prism 8. (E) Bisque deconvolution assigned relative cell type enrichment in bulk RNA samples from three wild-type and two N1ICDΔPEST-expressing kidneys using the top 20 or 100 markers from scRNA dataset (Combes et al., 2019) as a standard. Scale bar: 100 µm.
The rationale for this study stems in part from past practices of inferring fate outcomes from a limited number of ‘marker’ genes. To consider the broadest possible number of transcripts, we analyzed differential gene expression induced by stabilized N1ICDΔPEST in two N1ICDΔPEST overexpressing kidneys and three control littermates. mRNA isolated from each kidney anlage was submitted to Novogene for amplification, library construction and sequencing. 5283 transcripts were differentially expressed (Table S3). Among the 10% most underrepresented genes in the N1ICDΔPEST overexpressing kidney were Calb1 (LogFC −9; UB), Six2 (−8.7; NPCs), Slc12a1 and Slc12a3 (−8.26 and −8.16, respectively, DT/LOH), Sal3 (−6.6, NPCs, DT/LOH) and Fgf8 (−6, NPCs, DT/LOH). Tfap2b (−5, DT/LOH) and Sim1 (−4.7, DT) were in the next decile, ranking 56th and 65th, respectively. Hp (7.4, ePT) was present among the most over-represented decile, with Notch targets Cdh6 (1.6) and Nrarp (0.58) mildly overexpressed (Figs S5, S6 and Table S3). The near elimination of Six2 and other NPC genes (Fig. S6A) was consistent with the proposal that NICD repressed Six2 (Fujimura et al., 2010). To test whether the absence of Notch delayed niche exit and thus led to expansion of NPCs, we sorted GFP+ cells from kidneys isolated from four P0 Six2+/Cre; Notch2f/f and nine Six2+/Cre; Notch+/f pups. No significant difference was observed in the number of GFP+ NPCs (Fig. 5D), inconsistent with the proposed role for Notch as a Six2 repressor regulating niche exit. Notably, although Notch activation clearly can drive Six2-expressing cells to epithelialize, none of the known Notch targets were found in pretubular aggregate (PTA) in our (Jarmas et al., 2021) or in published (Combes et al., 2019) datasets, suggesting that it functions after the epithelia form, perhaps promoting proliferation (Fig. S7) and maturation.
Next, we manually annotated 2683 transcripts based on a published scRNA-seq dataset (Combes et al., 2019) to plot changes in gene expression as a heatmap segregated by assigned cell type (Fig. S6A and Table S3). NPC and DT/LOH genes were underrepresented in N1ICDΔPEST relative to control, while ePT, POD and inflammation markers were overrepresented in N1ICDΔPEST expressing kidneys. Notably, mature PT transcripts declined in N1ICDΔPEST overexpressing cells. To gain a more quantitative insight, we performed deconvolution analysis of the bulk data with BisqueMarker (Jew et al., 2020), applying a weighted PCA approach using either the top 20 or top 100 markers from Combes et al. (2019) (Fig. 5E, Table S5). Based on this analysis, PTA and DT were under-represented, while SSB and ePT were overrepresented in N1ICDΔPEST overexpressing kidneys. NPC, POD, RV and mature PT were underrepresented in the top 20 marker deconvolution and overrepresented using the top 100 markers, perhaps reflecting an abnormal redistribution of transcripts in these cells consistent with the supraphysiological presence of N1ICDΔPEST. In aggregate, results from manual curation and regression-based computational analyses are most consistent with a role for Notch in supporting selection of the proximal fates early, before promoting epithelial proliferation and maturation of both distal and proximal tubule cells.
NOTCH signaling required for proximal fates to emerge in human kidney organoid
To test the hypothesis that Notch function in nephron segmentation is conserved among mammals, we targeted the second exon of the human nicastrin (NCSTN) gene in the NHSK hiPSC line with gRNA and CRISPR-Cas9 (Fig. 6). Nicastrin is a non-redundant member of the γ-secretase complex and, in its absence, this enzyme is highly unstable, resulting in compete loss of signaling from all four Notch receptors (Li et al., 2003; Nguyen et al., 2006). GFP (and thus, CRISPR and gRNA)-expressing cells were sorted, and colonies screened for NCSTN expression and Notch activation (Fig. 6B,C). DNA from two NCSTN-deficient clonal cell lines was sequenced and found to contain deletions resulting in a frame shift and termination in exon 2 (clone 1) or 3 (Fig. 6A). Kidney organoids were generated from clone 20 (C20), an unmodified control from the same transfection, and C1 NCSTN−/− cells using a combined Morizane and Takasato protocols (see Materials and Methods; Morizane and Bonventre, 2017; Morizane et al., 2015). Organoids were analyzed for the expression of ePT, eDT and UB markers by confocal imaging, and z stacks were created for each. In contrast to robust nephrogenesis induced in the C20 organoid by day 22, C1 NCSTN−/− organoids were smaller and contained few clusters of KRT8/18-positive cells connected to TFAP2b-expressing tubules. PODXL-, HNF4A- or LTL-expressing cells, which are abundant in the parental line, were not detected (n=4 separate differentiation experiments, Fig. 6, Movie 3). This observation indicates that Notch function in proximal/distal axis segmentation is conserved across mammalian species.
CRISPR-CAS9-mediated NCSTN mutant iPSCs generate kidney organoids composed solely of immature distal tubular epithelia. (A) Schematic representation of NCSTN gene disruption by CRISPR/Cas9. The deleted nucleotides in exon 2 are indicated in blue and the resulting frameshifted amino acids and premature stop codon in the two mutated clones (1 and 14) are marked in red. (B) Western blot analysis for NCSTN expression in unmodified clones (C22 and C20) and NCSTN−/− clones (C1 and C14) showing loss of NCSNT protein expression in the latter. (C) NOTCH1 activation with Tryp/EDTA in an unmodified control (C20) and the two NCSTN−/− clones (C1 and C14) was analyzed by western blot probed with the αVal1744 rabbit monoclonal antibody, detecting the VLLS epitope of N1ICD. N1ICD was detected in C20 but not in either of the NCSTN−/− clones, indicting loss of γ-secretase. (D) Human iPSC-derived kidney organoids from C20 (unmodified control; lower panel) and C1 (NCSTN−/−; upper panel) show significantly lower percentages of differentiated structures in the NCSTN−/− compared with controls in bright-field images of kidney organoids derived from NCSTN−/− (a) and unmodified control (b) iPSC clones. (c-h) Whole-mount immunofluorescent staining for early proximal nephron structures (i.e. HNF4a, LTL and PODXL) and early distal nephron structures (i.e. TFAP2B and KRT8/18), demonstrating expression of distal nephron markers but not expression of proximal nephron markers in the NCSTN−/− (c,e,g), in contrast to unmodified control (d,f,h). (i,j) Magnified view of nephron epithelia shown in g,h, respectively. Both bright-field and immunofluorescent images are representative of four separate kidney organoid differentiation experiments for each cell clone. Eight to 10 organoids from every clone in each differentiation experiment were immunostained for these markers. Scale bars: 100 µm.
CRISPR-CAS9-mediated NCSTN mutant iPSCs generate kidney organoids composed solely of immature distal tubular epithelia. (A) Schematic representation of NCSTN gene disruption by CRISPR/Cas9. The deleted nucleotides in exon 2 are indicated in blue and the resulting frameshifted amino acids and premature stop codon in the two mutated clones (1 and 14) are marked in red. (B) Western blot analysis for NCSTN expression in unmodified clones (C22 and C20) and NCSTN−/− clones (C1 and C14) showing loss of NCSNT protein expression in the latter. (C) NOTCH1 activation with Tryp/EDTA in an unmodified control (C20) and the two NCSTN−/− clones (C1 and C14) was analyzed by western blot probed with the αVal1744 rabbit monoclonal antibody, detecting the VLLS epitope of N1ICD. N1ICD was detected in C20 but not in either of the NCSTN−/− clones, indicting loss of γ-secretase. (D) Human iPSC-derived kidney organoids from C20 (unmodified control; lower panel) and C1 (NCSTN−/−; upper panel) show significantly lower percentages of differentiated structures in the NCSTN−/− compared with controls in bright-field images of kidney organoids derived from NCSTN−/− (a) and unmodified control (b) iPSC clones. (c-h) Whole-mount immunofluorescent staining for early proximal nephron structures (i.e. HNF4a, LTL and PODXL) and early distal nephron structures (i.e. TFAP2B and KRT8/18), demonstrating expression of distal nephron markers but not expression of proximal nephron markers in the NCSTN−/− (c,e,g), in contrast to unmodified control (d,f,h). (i,j) Magnified view of nephron epithelia shown in g,h, respectively. Both bright-field and immunofluorescent images are representative of four separate kidney organoid differentiation experiments for each cell clone. Eight to 10 organoids from every clone in each differentiation experiment were immunostained for these markers. Scale bars: 100 µm.
DISCUSSION
Investigations into the developmental role of Notch signaling in the mammalian kidney began at the turn of the 21st century, with identification of a Jag1 mutation in Alagille syndrome (Heritage et al., 2000) and concurrent studies in model organisms (McCright et al., 2001). The importance of Notch signaling to nephrogenesis is broadly accepted, but a role in nephron segmentation is disputed. Here, we establish that, indeed, prior to its roles in maturation and proliferation, Notch signals contribute to the proximo-distal axis in the emerging nephron epithelia.
Notch pathway loss-of-function models demonstrated that MET can occur, albeit at reduced rates, and with a paucity of epithelial cells, but a definitive mechanistic explanation for this shortage was not provided. Notch promotes epithelial proliferation in a Jag1-dependent process (Cheng et al., 2007; Ungricht et al., 2022) and, perhaps, survival [one report noted increased apoptosis in the marginal zone mesenchyme of presenilin 1 (Psen1) hypomorphs at E15.5 (Wang et al., 2003)]. Interestingly, Jag1 promotes proliferation in trans (in neighboring cells) and inhibits it in cis (where it is expressed), the latter function only uncovered by screening chimeric organoids (Ungricht et al., 2022). There is also agreement that Notch signals likely stabilize and/or promote differentiation of all epithelial nephron fates, initially proposed by Fujimura et al. (2010) and elaborated by Chung et al. (2016, 2017). Notch is not likely to contribute to selection of differentiation versus self-renewal in the NPC: if Notch arbitrated this choice via lateral inhibition (Bray, 1998), Notch-deficient cells would be expected to assume an alternative fate. Accordingly, it was suggested that Notch inhibits Six2 (Chung et al., 2016; Fujimura et al., 2010), and that Six2-expressing NPCs accumulated in Notch-deficient kidneys (Chung et al., 2016). However, the same number of NPCs could occupy a larger area in the smaller Notch-deficient kidneys and, indeed, we show that the loss of Notch signaling did not alter the number of Six2+ progenitors (Fig. 5D). Whereas physiological Notch signals are not acting to switch cells from self-renewal to differentiation, Notch signals could promote proximal development within a developmental window, defined pharmacologically with γ-secretase inhibitors (Cheng et al., 2003, 2007). This role would predict that if Notch contributed to segmentation, in its absence, a skewed fate map will be observed. Testing this prediction produced results that varied based on the choice of markers used and can either support a role in segmentation (Cheng et al., 2007) or refute it (Chung et al., 2016, 2017). In addition, the broad domain in which Notch2 was deleted by the Pax3-Cre used in our study, the inability of N2ICD or of late expression of N1ICDΔPEST to promote proximal fate (Chung et al., 2016, 2017; Fujimura et al., 2010) and the absence of mature PT or DT markers after deleting Notch1/2 in RV epithelia (Chung et al., 2017) further diminished confidence in the proposed role for Notch in segmentation.
Single cell transcriptomics revealed that earlier studies addressing segmentation were deficient in selecting the markers for their analysis. Whereas we argued that the broad epithelial marker Cdh1 can be used to identify distal cells, we did not support that claim with any additional distal markers. The counterargument relied on the absence of other markers (Chung et al., 2017), which we now realize reflect the mature distal and proximal fate. A pan-epithelial block in maturation prevents expression of these markers, and their absence, does not exclude a Notch function in segmentation operating before maturation.
Our single cell transcriptional analysis of developing kidneys revealed a hierarchy of distal and proximal markers enabling a higher-resolution re-examination of Notch functions during nephrogenesis. Lineage analysis in two loss-of-function mouse models, Six2-Cre and Fgf20-Cre, and in a nicastrin-deleted human iPSC-derived kidney organoid model, reveals that these cells give rise to predominantly Tfap2b+ cells, a transcription factor committing cells to the distal/LOH fate, connected to KRT8/18-positive connecting duct cells (Marneros, 2020; Naganuma et al., 2021). We demonstrated that lower levels of or delayed Cre expression in mouse kidneys allows enough nephrons to form, permitting survival. Importantly, although urine analysis did not reveal any renal dysfunction, the proximal/distal ratio was altered in Fgf20-Cre, Notch2f/f kidneys. Based on this model, we anticipate that analysis of archival kidneys from Alagille patients will show a similar bias with fewer PT cells relative to DT/LOH.
Correspondingly, marker and transcriptome analyses of arrested anlage expressing a stabilized N1ICDΔPEST demonstrated a skewing towards immature early epithelia, no NPC, and few DT-fated cells. Notably, despite the strong bias towards the early proximal fate, Lhx1, which is thought to be a mediator of Notch activity (Chung et al., 2017), was among the transcripts suppressed by this hyper-physiological Notch signal, as was the Notch target gene Hes5 (Fig. S4A); Hnf1β levels were unchanged. The abundant Wt1+ cells, derived from Cre deletion (i.e. marked by the lineage tracer, as in Chung et al. (2017), were negative for GFP co-expressed with N1ICDΔPEST, consistent with in vitro observations (Boyle et al., 2011) and with the hypothesis that Notch hyperactivation is inconsistent with podocyte development (Cheng et al., 2007). Finally, and in contrast to the outcome when Notch2 (herein) or Notch2 and Notch1 genes are deleted with Wnt4-Cre (Chung et al., 2017), deletion of Rbpj using Wnt4-Cre had no impact on nephron development, suggesting that by the time the long-lived RBPJ protein has been depleted (Turkoz et al., 2016), all Notch functions in the NPC lineage have been fulfilled.
How can the various studies reviewed herein be reconciled into a single unifying model for Notch function? We arranged the published results along a developmental ruler (Fig. 7, x-axis) and sorted by outcome (y-axis, right) and mode of perturbation (y-axis, left). Analyses of Notch2 loss of function induced with either Pax3-Cre or Six2-Cre show that Notch2 was not required for proliferation of naïve or committed NPC, nor was it required for Wnt4 activation (Boyle et al., 2011; Wang et al., 2003). The cells expressing Notch2 and Cited1 are unlikely to encounter the very few ligand-expressing NPCs within the same domain (Fig. 1), and when they do, it may accelerate exit as we do not see marked NPC clones with the fate marker Notch2+/CreLO (Liu et al., 2013, Fig. 7). These observations in mice are supplemented by Network analysis in human first trimester kidneys (Fig. S5B; Lindstrom et al., 2021), consistent with Notch activity emerging after MET, in the PTA, where Notch signaling promotes early segmentation into ePT (it is dispensable in Tfap2b+ eDT). In all nephron epithelia, Notch signals are required for maturation, proliferation and/or survival (Chung et al., 2016, 2017; Fujimura et al., 2010; Ungricht et al., 2022) prior to or during the formation of the S-shape body (Rbpj deletion in this study and pharmacological analysis; Cheng et al., 2003). One consequence of the maturation-promoting Notch function is that cell fate choices in loss-of-function models cannot be analyzed with late differentiation-dependent lineage markers; this function can be detected by transcriptomic analysis or by the use of early markers (presented in Fig. 1, Fig. S1 and Naganuma et al., 2021). Deleting both receptors with Wnt4-Cre (acting in PTA/RV) is predicted to result in a similar bias, not detectable with the late markers used by Chung et al. (2017). By contrast, using the same Wnt4-Cre line to remove both Rbpj alleles falls outside the Notch-sensitive window due to the perdurance of the RBPJ protein (Fig. 7). When stabilized, hyper-physiological N1ICDΔPEST is overexpressed from the weak Rosa26 locus in the NPCs (using Six2-Cre or TAT-Cre), the pro-ePT differentiation function of Notch is obvious, but maturation is impaired by sustained strong Notch signals, as evident by the reduction in mature PT markers. As the NICD domains of Notch1 and Notch2 are interchangeable with no impact on kidney development (Liu et al., 2013), we favor the interpretation that lower amounts of PEST-containing N2ICD (Fujimura et al., 2010) are either insufficient to bias the outcome or do so in a manner requiring quantification of all transcripts.
The temporal requirements for Notch signaling revealed by genetic and pharmaceutical studies. Schematic representation of phenotypes along a developmental ruler (x-axis; IM, intermediate mesoderm, MM, metanephric mesenchyme, NPC, nephron progenitors, NPCc, committed NPC; PTA, pretubular aggregate, RV, renal vesicle, SSB, S-shape body) and sorted by outcome (y-axis, right) and mode of perturbation (y-axis, left). Below the developmental timeline is a representation for the expected expression onset of various Cre drivers (Pax3, Six2, Fgf20 and Wnt4). Of these drivers, Fgf20 expression levels are the lowest. The outcome column describes what is observed when conditional alleles, listed in the schematic, are removed. The triangular shape indicates time to complete loss of protein function. The tone of gray suggests wild type (N1+N2), reduced dose (N1 only) or no Notch signal (white). Graded red boxes depict the sensitivity to GSI (gamma-secretase inhibitor). Downward pointing arrows indicate fewer nephrons or DT cells than in control (the thickness of the arrow reflects the magnitude). Upward pointing arrow indicates more nephrons or PT cells than in control. The temporal window is outlined with a dashed red line. VSD, ventricular-sepal defect (asterisk; inferred from Caprioli et al. 2015). Glom, glomerulus. Inh., inhibitors. Exp., expression. Not included in the schema are the outcome in NCSTN-deleted organoids, which are similar to those in the presenilin 1 study by Wang et al. (2003). The sources of the data in the figure are: (1) Cheng et al. (2003); (2) Wang et al. (2003); (3) Cheng et al. (2007); (4) Fujimura et al. (2010); (5) Surendran et al. (2010a); (6) Chung et al. (2016); (7) Chung et al. (2017); and (8) this study.
The temporal requirements for Notch signaling revealed by genetic and pharmaceutical studies. Schematic representation of phenotypes along a developmental ruler (x-axis; IM, intermediate mesoderm, MM, metanephric mesenchyme, NPC, nephron progenitors, NPCc, committed NPC; PTA, pretubular aggregate, RV, renal vesicle, SSB, S-shape body) and sorted by outcome (y-axis, right) and mode of perturbation (y-axis, left). Below the developmental timeline is a representation for the expected expression onset of various Cre drivers (Pax3, Six2, Fgf20 and Wnt4). Of these drivers, Fgf20 expression levels are the lowest. The outcome column describes what is observed when conditional alleles, listed in the schematic, are removed. The triangular shape indicates time to complete loss of protein function. The tone of gray suggests wild type (N1+N2), reduced dose (N1 only) or no Notch signal (white). Graded red boxes depict the sensitivity to GSI (gamma-secretase inhibitor). Downward pointing arrows indicate fewer nephrons or DT cells than in control (the thickness of the arrow reflects the magnitude). Upward pointing arrow indicates more nephrons or PT cells than in control. The temporal window is outlined with a dashed red line. VSD, ventricular-sepal defect (asterisk; inferred from Caprioli et al. 2015). Glom, glomerulus. Inh., inhibitors. Exp., expression. Not included in the schema are the outcome in NCSTN-deleted organoids, which are similar to those in the presenilin 1 study by Wang et al. (2003). The sources of the data in the figure are: (1) Cheng et al. (2003); (2) Wang et al. (2003); (3) Cheng et al. (2007); (4) Fujimura et al. (2010); (5) Surendran et al. (2010a); (6) Chung et al. (2016); (7) Chung et al. (2017); and (8) this study.
In conclusion, this study integrates two decades of Notch research in the metanephric kidney context to refine a timeline in which three Notch functions operate: (1) assistance in selection of ePT fate; (2) promotion of proliferation/survival of nascent nephron epithelia; and (3) their maturation. We extend these conclusions to the human. It is unclear whether these Notch functions are separated by the S-phase, as they are during gonad development in C. elegans (Ambros, 1999), or by a cell division, as they are in the Drosophila peripheral sense organs (Jan and Jan, 1994). The strength of Notch signals is important, with weak/no early signal being compatible with the DT/LOH developmental trajectory but a somewhat stronger signal needed for ePT, such as the development of definitive hematopoietic stem cells in the AGM (Gama-Norton et al., 2015). All cells need an above-threshold signal to expand and mature. Too much (or persistent) signal, as in N1ICDΔPEST, is deleterious (see also Niranjan et al., 2008; Sweetwyne et al., 2015; Waters et al., 2008). It remains to be determined which downstream effectors are involved in prompting the ePT fate, nephron epithelial expansion and segment maturation.
MATERIALS AND METHODS
Mice
Mouse strains used were: Foxd1tm1(GFP/Cre)Amc (FoxD1-Cre) (Jackson Laboratory stock 012463); Six2tm3(EGFP/Cre/ERT2)Amc (Six2+/Cre; Jackson Laboratory stock 009600); Fgf20tm2.1(Cre/EGFP)Dor (Fgf20-Cre) (Huh et al., 2015); Notch2tm3grid (Notch2+/f) (Jackson Laboratory stock 010525); Rbpjtm1Hon (Rbpj+/f) (Tanigaki et al., 2002); Wnt4tm3(EGFP/Cre)Amc (Wnt4+/Cre) (Jackson Laboratory stock 032490); RosaNICD (Murtaugh et al., 2003); and Rosa+/EYFP [Gt(ROSA)26Sortm1(EYFP)Cos] (Srinivas et al., 1999).
Mice were maintained at the Cincinnati Children's Hospital Medical Center animal facility following animal care guidelines. Our experimental protocols (IACUC 2018-0108/0107) were approved by the Animal Studies Committee of CCHMC. For embryonic experiments, embryonic day (E) 0.5 was noted at noon at the day the copulatory plug was observed. Genotyping was performed on toe or tail clips following standard genotyping protocols (Stratman et al., 2003). Oligonucleotides used are listed in Table S6.
Hematoxylin and Eosin staining
Kidneys were isolated from mice at P0, E15.5, E17.5 and E18.5. Once isolated, P0 kidneys were fixed in 4% PFA for 16-20 h at 4°C and younger kidneys were fixed for 1 h at room temperature. After washing three times in PBS, kidneys were processed, embedded and sectioned by the CCHMC Pathology Core. Slides were then deparaffinized in xylene three times for 2 min each and rehydrated in ethanol (100%, 70%, 50% and 30%). Slides were then rinsed with distilled water. After staining with Hematoxylin for 4-5 min, slides were rinsed in H2O, dipped once in 0.05% acetic acid, rinsed in H2O, dipped in 0.1% ammonium hydroxide three times, rinsed in H2O and dipped in 80% ethanol 20 times. Once stable in 80% ethanol, they were dipped in Eosin for 20 s, rinsed in an ethanol series to xylene before being mounted with mounting medium. Slides were left to dry overnight before being imaged on a Nikon NiE upright microscope.
Immunofluorescent staining
Kidneys were isolated and fixed as above. After washing three times in PBS, kidneys were processed in an alcohol series to xylene, embedded in paraffin wax and sectioned by the CCHMC Pathology Core. Slides were deparaffinized in xylene twice for 10 min each and rehydrated as above. Slides were then washed in PBS three times for 5 min each, placed in a solution of Trilogy Antigen Retrieval (Cell Marque, 920P-10) and boiled in a rice cooker for 30 min. Slides were washed again in PBS three times for 5 min each and then placed in a blocking solution [PBS, 5% bovine serum albumin (BSA; Sigma-Aldrich, A8806-1G), 0.1% Tween-20 (Fisher Scientific, BP337-500) and 10% normal donkey serum (NDS; Jackson Immunoresearch Laboratories, 017-000-121)] for 1 h at room temperature. See Table S5 for the list of antibodies used. Images were captured using a Nikon A1 inverted confocal microscope. Slides were incubated in secondary antibody for 1 h at room temperature. Slides were then washed a final three in PBS-T for 30 min each, and mounted with Prolong Gold (Cell Signaling Technology, 9071S).
Whole mount preparation
Kidneys were removed from NICD mice at E15.5 and fixed in 4% PFA for 1 h at room temperature, washed twice with PBS and permeabilized and blocked for 3 h with gentle rocking at room temperature in TSP-NS [PBS, 0.1% Triton 100X (Fisher Scientific, BP151-500), 0.05% saponin (Fluka, 47036) and 10% normal donkey serum]. Whole kidneys were submerged overnight in TSP-NS with specified primary antibodies (Table S5) at 4°C with gentle rocking. Post-incubation, the kidneys were washed three times for 10 min each in TSP at room temperature. Secondary antibodies were diluted in TSP-NS (Table S5) and placed onto kidneys for a 3 h incubation at room temperature with gentle rocking. Kidneys were then washed three times with TSP for 15 min each, placed into PBS and covered with aluminum foil for storage. The day before imaging on a confocal microscope, the tissue was cleared in Refractive Index Matching Solution [RIMS; 40 g Histodenz (Sigma D2158)/0.02 M phosphate buffer, 0.01% sodium azide, 0.1% Tween 20 and 1 g DABCO] for 3-5 h, after which the solution was replaced with fresh RIMS. Tissue was then placed into a 35 mm dish (No. 1.5 Uncoated Coverslip, 20 mm Glass Diameter; MatTek, P35G-1.5-20-C.S) and covered with enough RIMS to avoid an air bubble. Images were captured using a Nikon A1 Inverted confocal microscope.
RNA extraction
Kidneys were isolated from E15.5 mice and placed into ice-cold PBS. DNA was collected from tails of embryos for genotyping purposes. RNA was extracted using the Invitrogen PureLink RNA Mini Kit (Thermo Fisher Scientific, 12183025) following the manufacturer's protocol. Samples were then sent to Novogene for RNA sequencing.
Cell type assignment and deconvolution of bulk RNA
To assign cell types manually, we used the VLOOKUP table in Table S4 (Combes et al. 2019) and entered our differentially expressed, filtered bulk gene list (FDR >0.05, −0.5>LogFC>0.5). After all the expression values in NPC were computed by the excel VLOOKUP script, we sorted by imputed cell type and annotated the gene based on the cell type with unique or highest expression value if LogFC was >0.65. To estimate cell type proportions in bulk RNA-Sequencing data while considering their expression levels in the bulk dataset and the reference data, we performed reference-based decomposition with the R package BisqueMarker (Jew et al., 2020). Briefly, expressed genes with zero variance were filtered out and the remaining expressed gene counts were converted to counts per million. Markers for each cell type from single cell dataset are filtered at FDR <0.5 and are then ranked based on P-value (Combes et al., 2019). To estimate cell type proportions in bulk samples, Bisque applies a weighted principal component analysis (PCA) approach, where the 1st principal component is calculated on the expression matrix using a subset of the marker genes. Expression matrix of markers are scaled and each gene column is multiplied by its weight (log-fold change). We chose the Combes dataset over our own (Jarmas et al., 2021) because, at the time of this analysis, our data were yet to complete peer-review, and the Combes dataset included all cell types, whereas ours was cortically biased.
Quantification
Quantification of distal and proximal tubule markers were carried out using NIS Elements with help from Dr Matt Kofron. Binaries were created to measure the total kidney area and the areas containing distal markers, proximal markers or GFP (in μm2). The binaries were kept the same for each slide that was imaged, and modified only to account for GFP gene dose.
Renal panel
Littermates were euthanized between day 60 and day 90 for renal panels. Blood was removed within 5 min of being euthanized to avoid clotting, placed into serum separator tubes (Fisher Scientific, 02-675-185) and centrifuged at 21,130 g for 5 min. Blood samples were then sent to IDEXX Laboratories to be tested for renal function. The kidneys were also removed at this time for nephron counts.
Nephron counts
HCl maceration of whole kidneys was performed as described previously (MacKay et al., 1987; Peterson et al., 2019). Kidneys isolated from >P28 animals were minced using a razor blade after removal of the capsule, and incubated in the presence of 6 N HCl for 90 min. The dissociated kidneys were vigorously pipetted every 30 min to further disrupt them. After incubation, five volumes of distilled water were added to the samples followed by incubation at 4°C overnight. 100 μl of the thoroughly mixed macerate was then pipetted into a cell culture dish and glomeruli were counted in triplicate (three aliquots) for each sample. A single experimenter, blinded to the genotypes of the kidneys being scored, performed all counts, to avoid interobserver variability.
Flow cytometry
Briefly, P0 kidneys were isolated from Six2+/Cre; Notch2+/f and Six2+/Cre; Notch2f/f littermates, the capsule removed and the kidneys digested in 250 μl of 1 mgl/ml Collagenase D shaking in an Eppendorf Thermomixer R at 1400 rpm for 10-15 min at 37°C to remove the peripheral nephron progenitor cells. The remnant kidney tissue was removed and the cell suspension was washed twice with 1% BSA/PBS. The number of GFP-expressing cells was quantified by a Sony SH800S flow cytometer (Sony Biotechnology).
CRISPR/CAS9-mediated deletion of NCSTN in the NHSK iPSC line
CRISPR-Cas9 was used to introduce a mutation in an iPSC line (NHSK) derived from skin cells of a female donor with a normal karyotype. Two guide RNAs (gRNAs) were designed to induce a stop codon in exon 2 of the NCSTN gene (AAACTCTCAACTCTCACTGGCAGCC and CACCGGCTGCCAGTGTGAGTTGAAG) (Fig. 6A). The gRNA target sequences were selected according to the on- and off-target scores from http://CRISPOR.org web tool (Haeussler et al., 2016) and cloned into the pX458M-HF vector (modified from the pX458 vector, Addgene 48138) to carry an optimized sgRNA scaffold and a high-fidelity eSpCas9(1.1)-2A-GFP (Chen et al., 2013; Slaymaker et al., 2016). The editing activity of the plasmid was validated in 293T cells by the T7E1 assay. A single cell suspension of iPSCs was prepared using accutase and 1×10−6 cells were nucleofected with 5 μg of the plasmid using program CA137. Forty-eight hours post-nucleofection, GFP-positive cells were isolated by FACS and replated at cloning density in hESC media containing 20% Knockout Serum Replacement (KOSR) (Gibco), 4 ng/ml bFGF and 10 μM Y27632 (inhibitor of Rho-associated, coiled-coil containing protein kinase; ROCK) in six-well dishes containing 187,500/well mitomycin C-inactivated CF1 MEFs. After 1-2 weeks, single clones with stereotypical iPSC morphology were manually excised and transferred to mTeSR1/Matrigel culture conditions for genotyping, expansion and cryopreservation. PCR and enzyme digestion identified candidates for correctly targeted clones. DNA from two NCSTN-deficient clones as well as an unmodified control (C20) and parental clone (NHSK) was PCR amplified and sequenced to establish the molecular nature of the mutation. NCSTN−/− (C1) and unmodified control (C20) expanded clones were analyzed by western blot for NCSTN expression with an anti-nicastrin antibody (N1660, Sigma-Aldrich).
NOTCH1 activation with EDTA was performed on NCSTN−/− (C1 and C14) expanded clones and parental NHSK iPSCs as previously described (Ilagan et al., 2011; Rand et al., 2000). Briefly, cells were washed with phosphate-buffered saline solution (PBS; Gibco) and then incubated in Trypsin/EDTA (Invitrogen) for 10 min at 37°C. Following cell detachment, mTeSR medium was added to each well and cells were incubated at 37°C for 2 h. Cells were then collected and western blot analysis was performed with cleaved Notch1 (Val1744) rabbit monoclonal antibody (Stem Cell Biologies).
Generation of iPSC-derived 3D kidney organoid
Human iPSCs (either C20 or C1) were grown on six-well plates covered with Matrigel (MG) in defined, feeder-free maintenance medium (mTeSR for human ES and iPS cells, Stem Cell Technologies, 85850) as previously described (Low et al., 2019; Yu et al., 2007). When cells reached ∼80% confluency (day −1), medium was replaced with mTeSR containing 10 mM of Rock inhibitor. Differentiation into 3D kidney organoids was performed according to the Morizane protocol (Morizane and Bonventre, 2017) with some modifications. Briefly, on day 0 of the differentiation protocol, cells were dissociated via Accutase (STEMCELL Technologies) and plated in 24-well plates covered with matrigel at 20,000 cells/per well and grown in basic differentiation media (advanced RPMI with Glutamax) with sequential addition of CHIR (8 µM) and noggin (days 0-3), followed by activin A (days 4-6) and FGF9 (10 ng/ml, days 7-9), as previously described (Morizane and Bonventre, 2017), with daily media changes until day 9. On day 9 of differentiation, cells were dissociated with Accutase followed by aggregation on transwell membranes at an air/liquid interphase, as previously described (Takasato et al., 2016). Basic differentiation medium supplemented with CHIR (3 µM) and FGF9 (10 ng/ml) was added to the bottom of the transwell. The cell aggregates were then cultured at 37°C in 5% CO2 with daily media changes for an additional 12-14 days. On day 9+2, the medium was changed to the basic differentiation medium supplemented with FGF9 10 ng/ml (days 9+2 to 9+4). After that, the organoids were cultured in basic differentiation medium with no additional factors (days 9+5 to 9+12). Total number of days to analysis was 21-24.
Whole-mount immunofluorescence of 3D kidney organoids
Organoids were fixed in transwell plates with 4% PFA at 4°C for 20 min as previously described (Takasato et al., 2015). 4% PFA was removed and organoids were washed three times with Dulbecco's phosphate-buffered saline (DPBS, Gibco). 150 μl of blocking buffer (10% donkey serum/0.3% Triton X-100/DPBS) was added into each well of a 24 well plate. Organoids were cut off the filter and submerged in the blocking buffer at room temperature for 2-3 h on a rocker. Primary antibodies were prepared in blocking buffer (0.3% Triton X/10% donkey serum/DPBS) with appropriate dilutions (Table S5). Blocking buffer was aspirated off the 24-well plate, and 150 μl of primary antibody solution was added into each well. Organoids with primary antibody were incubated at 4°C overnight while shaking and protecting from light. Primary antibody solution was removed from each plate and organoids were washed with PBTX (0.3% Triton X-100 and DPBS). Appropriate secondary antibodies (all 1:400 dilution, Table S5) were diluted in PBTX and 150 μl was added into each well in the 24-well plate.
Organoids were incubated in secondary antibody solution at 4°C overnight while shaking and protecting from light. Following removal of secondary antibody solution, organoids were either incubated with 20 mg/ml DAPI (1:1000 dilution) in DPBS for 3 h and then washed, or washed with DPBS. Slides were mounted with Prolong Gold (Cell Signaling Technology, 9071S) and kept at 4°C overnight until visualized on a Nikon A1 Inverted confocal microscope.
Acknowledgements
We thank Dr Eric Brunskill for encouragement, advice and for the creation of the NCSTN-deleted iPSC line; Dr Yueh-Chiang Hu and the Transgenic Animal and Genome Editing Core Facility; the Research Flow Cytometry Core; Dr Chris Mayhew and the Pluripotent Stem Cell Facility; and the Confocal Imaging Core at Cincinnati Children's Hospital Medical Center. The Research Flow Cytometry Core is supported by NIH S10OD023410. We thank Drs Meredith Schuh, Aaron Zorn and Joo-Seop Park for critical comments.
Footnotes
Author contributions
Conceptualization: R.K.; Methodology: K.D., N.P.S., R.K.; Validation: K.D., L.C., P.C.; Formal analysis: K.D., L.C., A.J.P., N.P.S., P.C.; Investigation: K.D., L.C., A.J.P., N.P.S.; Resources: R.K.; Data curation: K.D., L.C., R.K.; Writing - original draft: K.D., R.K.; Writing - review & editing: K.D., L.C., A.J.P., N.P.S., P.C., R.K.; Supervision: R.K.; Project administration: R.K.; Funding acquisition: R.K.
Funding
This study was supported by funds from the William K. Schubert Endowment to R.K., by the National Institute of Diabetes and Digestive and Kidney Diseases (NIH DK106225 to R.K.; 1F30 DK 123841 to A.J.P.) as well as generous financial support for the cores at Cincinnati Children's Hospital Medical Center. Deposited in PMC for release after 12 months.
Data availability
The scRNA-sequencing data analyzed in this study was published by Jarmas et al. (2021) and is deposited in GEO under accession codes GSE173264, GSE173265 and GSE173266, and contain all processed bam files, raw counts of genes across barcodes/cells and cell annotations in the form of a metafile. The raw and processed data discussed in Fig. 5 have been deposited in GEO under accession number GSE201938.
Peer review history
The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.200446.
References
Competing interests
The authors declare no competing or financial interests.