Endogenous Notch Signaling in Adult Kidneys Maintains Segment-Specific Epithelial Cell Types of the Distal Tubules and Collecting Ducts to Ensure Water Homeostasis

Induction of Dominant-Negative Mastermind-Like-1 Expression in Adult Renal Epithelia Alters the Cell Type Composition of Kidney Collecting Ducts and Diminishes Urine Concentrating Capacity

Adult mouse kidney collecting ducts express the Notch ligand Delta-like-1 (Dll1) in ICs adjacent to Aqp3⁺ PCs (Figure 1A). Consistent with active Notch signaling in adult collecting ducts, we observed H2B::Venus⁺ PCs in adult Notch reporter mice³⁰ (Figure 1B). To determine if Notch signaling is required for maintenance of mature collecting duct cell types, expression of dominant-negative mastermind-like-1 (dnMaml), an inhibitor of Notch signaling–mediated transcription,³¹ was induced in nephrons and collecting ducts of 3- to 4-week-old Pax8->rtTA;TRE->dnMaml mice by dox treatment.^32–34 Consistent with inhibition of Notch signaling, the expression level of Nrarp, a target of Notch signaling,^35,36 is reduced after dox treatment (Figure 1C). Induction of dnMaml reduced expression levels of PC genes Aqp2, Aqp4, and Avpr2, and increased levels of IC-specific genes Foxi1 and Slc4a9 (Figure 1C). Consistent with altered intercalated and PC-specific gene expression levels, the number of CAII⁺ ICs increased and Aqp2⁺ PCs decreased with dnMaml expression (Figure 1D). Along with changes in collecting duct cell types and gene expression levels, we detected reduced 24-hour urine osmolality in mice with dnMaml induction as early as 3 days after dox treatment compared with dox-treated controls (Figure 1F).

Inactivation of Notch1 and Notch2, or Hes1, in Adult Kidneys Triggers Loss of PCs and Gain of Intercalated Cell Types

Although dnMaml blocks Notch signaling–mediated transcription,³¹ it is possible that the adult kidney phenotype with induction of dnMaml operates through a mechanism independent of blocking Notch signaling. To confirm the requirement of Notch signaling in adult kidneys, floxed alleles of Notch1 and Notch2 (N1f/f;N2f/f) were inactivated using Pax8->rtTA;TRE->Cre system in 3-week-old mouse nephrons and collecting ducts. Dox treatment of Pax8->rtTA;TRE->Cre;N1f/f; N2*/f mice (Notch1 & 2 cKO), where * denotes floxed or deleted allele, from 3 to 5 weeks of age resulted in a severe urine concentrating deficit by 8 weeks of age (n=6 controls; n=9 Notch1 & 2 cKO; Figure 2H). Because Notch signaling is critical for nephron formation and proximal tubule specification during kidney development,^37,38 we examined the proximal tubules in Notch1 & 2cKO mice, but did not observe any obvious morphologic changes or significant changes in the expression levels of proximal tubule specific genes, such as megalin, in the kidneys of Notch1 & 2 cKO 8-week-old mice (data not shown). Instead, we observed that the Notch1 & 2 cKO kidneys have increased Atp6v1b1⁺ ICs in the CCDs, inner stripes of OMCDs, and in the initial IMCDs compared with that of control kidneys (Figure 2, A–F and I). The increase in ICs is further supported by increased mRNA levels of Foxi1 and Atp6v1b1 in the whole Notch1 & 2 cKO kidneys compared with wild-type controls (Figure 2G). The increase in ICs accompanies a decrease in PCs, as evidenced by fewer Aqp2⁺ cells (Figure 2, A–F and I). All collecting duct segments had fewer PCs in Notch1 & 2 cKO kidneys, except for those in the papillary regions (inner two thirds of the inner medulla), which were not noticeably different from control kidneys (data not shown). This does not appear to be due to inefficient inactivation in the papillary region as we detected very few papillary collecting duct cells with Notch2 mRNA in Notch1 & 2 cKO kidneys compared with wild-type controls (Supplemental Figure 2). The decrease in PCs was further confirmed using Aqp4 as a PC marker (Supplemental Figure 3) and by the overall reduction in Aqp2 mRNA levels (Figure 2G). Next, we stained for markers of mitochondria as they are more abundant in ICs than in PCs. Staining for mitochondrial markers CoxIV and Cytochrome C (Supplemental Figure 3) revealed their abundance within Aqp4-negative ICs and not as much in Aqp4⁺ PCs in control kidneys. In Notch1 & 2 cKO kidneys, most of Aqp4-negative duct cells expressed abundant amounts of mitochondrial markers, and most Aqp4⁺ duct cells had very few mitochondria (Supplemental Figure 3). TEM of Notch1 & 2 cKO kidney sections reveals that the increased number of ICs contain many more mitochondria compared with the PCs in the inner stripes of OMCDs (Supplemental Figure 4).

Hes1 is among the earliest genes upregulated by ectopic activated-Notch1 expression in the developing mouse collecting ducts,²³ and increased Hes1 staining is observed in PCs of SPAK kinase–deficient adult mouse kidneys.³⁹ Inactivation of Hes1 in adult mouse collecting ducts and nephrons by providing dox to Pax8->rtTA; TRE->Cre; Hes1f/f (Hes1cKO) mice from 3 to 5 weeks of age resulted in a severe urine concentrating deficit (Figure 3H), an increase in Atp6v1b1-expressing ICs, and a decrease in Aqp2-expressing PCs (Figure 3, A–F, and I). Consistent with the change in cell types, quantitative RT-PCR using RNA extracts from whole kidneys revealed increased mRNA levels of Foxi1 and Atp6v1b1, and decreased levels of Aqp2 in Hes1cKO kidneys by 1 week after dox treatment (Figure 3).

Segment-Specific Intercalated Cell Types Replace the Aqp2⁺ Cell Types after Notch Signaling Inhibition in Collecting Ducts and Distal Tubules

ICs are more abundant in the adult kidney collecting ducts after genetic inactivation of Notch signaling. We examined if loss of Notch increases a specific type (A, B, or non-A-non-B) of IC. The type A IC, identified by basolateral expression of the chloride bicarbonate exchanger AE1 and apical expression of the H⁺-ATPase pump, is the predominant IC type present in the outer and inner medulla, but not cortex of wild-type kidneys (Figure 4, A, D, and G). Interestingly, it is predominantly the type A IC that is increased in the Notch signaling–deficient OMCD and IMCD segments (Figure 4, E, F, H, and I). Both type B and type non-A-non-B ICs do not express AE1, whereas H⁺-ATPase pumps are on the basolateral membrane of type B ICs, and in diffuse vesicles in type non-A-non-B ICs. The type B and non-A-non-B ICs are present predominantly in the cortex of wild-type kidneys (Figure 4A) and are present in increased numbers in the cortex of Notch signaling–deficient cortical tubules (Figure 4, B and C). Pendrin, another marker of type B and non-A-non-B ICs, is mainly present in cortical ICs of wild-type kidneys, and increased in cortical cells of Notch signaling–deficient kidneys (Figure 4, J and K). We did not observe pendrin-positive cells in the OMCDs and IMCDs in the Notch signaling–deficient kidneys (data not shown). Consistent with Notch signaling in adult kidneys preventing expansion of intercalated cell types expressing pendrin (type B and type non-A-non-B) and those expressing AE1 (type A), there is an overall increase in the mRNA levels of Slc4a1 and Slc26a4 in the Hes1cKO kidneys 1 week after initiation of Hes1 inactivation (Figure 4L). This trend continues for Slc26a4 in the Notch1 & 2cKO kidneys even 5 weeks after initiation of Notch1 and Notch2 inactivation (Figure 4L).

Unlike the collecting ducts, the CNT of the distal nephron, which connects each nephron to the branched collecting duct network, consists of ICs intermingled with a CNT cell type that expresses both Aqp2 and calbindin-D28k (Calb1).³⁹ Considering that the CNT, along with the rest of the nephron, originate from a lineage distinct from the collecting ducts,^40,41 we examined whether maintenance of Calb1⁺; Aqp2⁺ cells in the CNT is also dependent on Notch signaling. Interestingly, the number of Aqp2⁺ and Calb1⁺ CNT cells and the number of Aqp2⁻ and Calb1⁺ DCT2 cells were decreased in the Notch signaling–deficient kidneys along with an increase in Atp6v1b1⁺ ICs in these segments (Figure 5D). In wild-type CNTs, each IC is surrounded by Aqp2⁺ and Calb1⁺ cells (arrowheads, Figure 5A) whereas in Notch signaling–deficient CNTs, there are many regions containing two or more ICs next to each other (asterisks in Figure 5, B and C) and a reduction in number of Aqp2⁺ and Calb1⁺ cells from approximately 65% in wild-type kidneys to about 30% in Notch signaling–deficient kidneys (Figure 5D). The Aqp2⁻ and Calb1⁺ DCT2 cells normally comprise 75% of the DCT2 segment, but are reduced to 40%–50% in the Notch signaling–deficient DCT2 segments (Figure 5D). These observations reveal the requirement for Notch1, Notch2, and Hes1 in the maintenance of Aqp2⁺ and Calb1⁺, as well as Aqp2⁻ and Calb1⁺ epithelial cell types in the distal tubule segments of adult nephrons.

Fate Tracing of Mature PCs in the Inner Stripe of the Outer Medulla Reveals Conversion to Intercalated Cells after Inactivation of Hes1 and, to Lesser Extent, Lithium Treatment

Loss of Notch signaling in adult kidneys results in increased ICs at the expense of PCs. Although there are several possibilities for increased ICs, one intriguing hypothesis is that PCs convert to ICs. The Elf5⁺ PC lineage-tracing tool previously described was used to test for direct conversion between cell types.²³ In this system, Elf5 drives rtTA and GFP expression (Elf5->rtTA;Tet-O-Cre; Rosa^+/tdtomato), and Cre expression in the presence of dox recombines Lox-P sites flanking a transcriptional stop placed upstream of tdtomato coding region at the Rosa locus (Rosa^+/tdtomato). We have previously observed that mice with Tet-O-Cre;Rosa^+/tdtomato have a subset of renal interstitial cells labeled with tdtomato independently of inheriting the Elf5->rtTA transgene. Importantly, only the PCs in the tubular epithelial structures of Elf5->rtTA;Tet-O-Cre; Rosa^+/tdtomato mice are labeled in a dox-dependent manner, and dox administration does not label renal epithelial cells in Tet-O-Cre; Rosa^+/tdtomato mice.²³ The Elf5⁺ PCs were labeled with tdtomato in adult mouse kidney collecting ducts by administering dox from 3 to 5 weeks of age (Figure 6A), along with or without inactivation of Hes1 in Elf5-expressing cells, and the fate of tdtomato-labeled PCs was determined at 6 weeks of age. The pattern of tdtomato-labeled collecting duct cells was similar to the expression pattern of Elf5 in adult kidneys,²³ with the highest labeling in papillary regions, moderate labeling in medullary regions, and minimal labeling in the cortex. When kidneys were stained for either PC marker Aqp4 (Figure 6, B and C) or IC markers c-Kit²² or Atp6v1b1 (Figure 6, D–G), most (approximately 97%) tdtomato⁺ OMCD cells were Aqp4⁺ (arrowheads in Figure 6B) and almost none (<0.2%) were Atp6v1b1⁺ or c-Kit⁺ (arrowheads in Figure 6, D and F) in wild-type mice 3–5 weeks after initiation of PC labeling. In contrast, a significant number of tdtomato⁺ duct cells begin expressing IC markers 3 weeks after initiation of Hes1 inactivation in Elf5-expressing cells (arrows in Figure 6, E and G), with an average of approximately 40% of tdtomato⁺ OMCD cells expressing IC markers (Figure 6J). An average of 29% of tdtomato⁺ OMCD cells expressing c-Kit still have Aqp4 lingering in the basolateral membranes in Hes1 mutant kidneys (Supplemental Figure 5), whereas 0% of the tdtomato⁺ OMCD cells express both c-Kit and Aqp4 in wild-type kidneys (n=4 control, n=3 Hes1 mutants, at least 100 cells analyzed per mouse). To determine whether the tdtomato⁺ ICs in the Elf5->rtTA;Tet-O-Cre; Rosa^+/tdtomato; Hes1^f/f mice arise from cells that undergo DNA synthesis and proliferation, we provided BrdU in the drinking water from time of dox treatment to tissue harvest (Figure 6A). Whereas intestinal epithelial cells had BrdU labeling (Supplemental Figure 5A), none of the c-Kit⁺, Aqp4⁺, or tdtomato⁺ collecting duct cells in the Elf5->rtTA;Tet-O-Cre; Rosa^+/tdtomato;Hes1^f/f mice were BrdU-labeled (Supplemental Figure 5). The absence of BrdU labeling within the collecting duct cells implies that they have not undergone DNA synthesis since the initiation of Hes1 inactivation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Inhibition of Notch signaling in Elf5⁺ mature PCs triggers a significant increase in principal-to-intercalated cell conversion in the inner stripes of the OMCDs. (A) Dox treatment from 3 to 5 weeks of age pulse-labels a subset of PCs in Elf5->rtTA; Tet-O-Cre; Rosa^+/tdtomato mouse kidneys by turning on expression of tdtomato. (B, D, and F) In wild-type OMCD, tdtomato labeling occurs predominantly in cells expressing Aqp4 (arrowheads in B) and tdtomato labeling almost never occurs in intercalated cells expressing Atp6v1b1 (asterisks in D) or c-Kit (asterisks in F). (C, E, and G) Notch signaling inhibition by inducible inactivation of Hes1 in Elf5-expressing PCs shows that some tdtomato-labeled OMCD cells do not express Aqp4 (arrow in C), whereas many other tdtomato-labeled cells still express Aqp4 (arrowheads in C). In the absence of Notch signaling, many of the tdtomato-labeled OMCD cells express intercalated markers Atp6v1b1 (arrows in E) or c-Kit (arrows in G). (A’) Schematic of lithium treatment. Dox was administered from 3 to 5 weeks of age to label PCs with tdtomato. Lithium was provided in diet from 6 to 8 weeks of age. Urine was collected before and after lithium treatment. (H and I) Fate tracing of PCs in three Elf5->rtTA;TRE->Cre; Rosa^+/tdtomato mice treated with lithium for 2 weeks revealed that the labeled cells still mostly remain principal and express Aqp4 (arrowheads in H). However, about 6% of the labeled cells begin expressing intercalated markers, such as Atp6v1b1 (arrows in I). The white dashes in (D, E, and I) outline the borders of OMCDs. (J) Graph showing that about 40% of tdtomato labeled cells in Hes1-deficient inner stripes of the OMCDs begin expressing intercalated cell markers, whereas 6% do so after 2 weeks of lithium treatment and <0.2% do so in wild-type kidneys. *Statistically significant difference between groups as determined by one-way ANOVA (F=25.16; P<4×10⁻⁴), and comparing three groups consisting of four wild-type, three Hes1-deficient mice, and three lithium-treated mice. Post hoc analysis pairwise comparison using the Tukey method revealed significant differences between wild-type and Hes-1-deficient groups (adjusted P<2×10⁻²), as well as between lithium-treated and Hes1-deficient groups (adjusted P<4×10⁻⁴). Scale bars, 10 µm in (B–I).

Considering that lithium treatment has been associated with remodeling of the adult collecting ducts and that it is hypothesized that lithium-induced NDI involves PC-to-IC conversion, we used Elf5->rtTA;Tet-O-Cre; Rosa^+/tdtomato mice to trace the fate of Elf5⁺ OMCD PCs after 2 weeks of lithium treatment (Figure 6A’). Lithium treatment caused a marked drop in urine concentrating capacity and increased number of tdtomato-labeled Atp6v1b1⁺ ICs compared with nonlithium-treated mouse kidneys (data not shown). Two weeks of lithium on average triggers 6% of the tdtomato-labeled PCs in the OMCD to begin expressing IC-specific markers (Figure 6I), which is an increase in PC-to-IC conversion, but is not of statistically significant difference when compared with <0.2% in nonlithium-treated wild-type mice (Figure 6J).

The PC-to-IC Conversion Triggered by Notch Signaling Inactivation in the Adult Kidneys Involves Intermediary Cell States and Upregulation of Foxi1

We next provided BrdU in the drinking water during 2 weeks of dox treatment to 3-week-old Pax8->rtTA;TRE->Cre;Hes1f/f mice to determine if cell type conversion required proliferation. Two weeks after initiation of Hes1 inactivation, the OMCD composition changes from 77% PCs and 23% ICs to 49% PCs, 41% ICs, and 6% PCs and ICs (Figure 7, A–C). Interestingly, although there is almost no incorporation of BrdU in control littermate OMCD cells (average of 0.085%; Figure 7A), 13.8% of OMCD cells are BrdU-labeled with Hes1 inactivation, of which 10.5% are Aqp4⁺, 2.3% are c-Kit⁺, 0.4% are Aqp4⁺ and c-Kit⁺, and 0.7% are Aqp4⁻ and c-Kit⁻ (Figure 7, B and C). About 95% of the c-Kit⁺ OMCD cells are not BrdU-labeled during 2 weeks after initiation of Hes1 inactivation, as evidenced by rows of ICs expressing c-Kit that are not BrdU-labeled (Figure 7B). Hence, direct conversion of PC to IC occurs without requiring DNA synthesis. In an attempt to capture the direct conversion, we stained for Aqp4 and Foxi1 using 100-μm-thick kidney sections from Hes1cKO mice 1 week after initiation of Hes1 inactivation and from control littermates. Maximum intensity projections (Figure 7, D and E) and 3D reconstructions (Figure 7, F and G, Supplemental Movies 1 and 2) from Z-stack confocal images revealed OMCD cells in the Hes1cKO and not in wild-type controls that express both Foxi1 in the nucleus and Aqp4 on the basolateral membrane (asterisks in Figure 7, D–G). Quantification of each cell type per tubule length revealed a significant average increase in intermediate cells (Foxi1⁺ and Aqp4⁺) along with a significant average reduction of Aqp4⁺ PCs in Hes1cKO compared with wild-type mice (Figure 7H).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Principal-to-intercalated cell conversion triggered by inactivation of Hes1 involves intermediary cell states and upregulation of Foxi1. (A–C) Kidneys from Hes1cKO mice given BrdU in the drinking water during 2 weeks of dox-mediated inactivation of Hes1 show an increase in the number of BrdU-labeled PCs from 0.1% to about 10% and BrdU-labeled intercalated cells from 0% to 2% (A–C). A majority of the increase in intercalated cells in the inner stripe of the OMCDs from 20% in wild-type to 40% in Hes1cKO mice (n=4 per genotype) does not involve BrdU-labeled cells (asterisks in B). In (C), * indicates significant differences by two-tailed unpaired t test when comparing percentages of Aqp4⁺ (P=6.5×10⁻⁵), cKit⁺ (P=0.001753), Aqp4⁺ and cKit⁺ (P=0.007), Aqp4⁻ and cKit⁻ (P=0.009), BrdU⁺ (P=0.04), Aqp4⁺ and BrdU⁺ (P=0.04), or cKit⁺ and BrdU⁺ (P=0.03) OMCD cells between wild-type and Hes1cKO mouse kidneys after 2 weeks of dox treatment (n=4 mice per genotype with at least 1400 OMCD cells analyzed per mouse). (D and E) Maximum intensity projections of confocal Z-stack images (optical sections) captured every 0.175 μm, using a ×60 objective on a Nikon A1 confocal microscope. (F and G) 3D ISO surface images of inner stripes of the OMCD segments generated from Z-stack images using Arivis Vision 4D. (D–G) Different views of a wild-type OMCD segment (D and F) compared with different views of a Hes1cKO OMCD segment (E and G) reveal Foxi1-expressing intercalated cells (arrowheads), Aqp4-expressing PCs (arrows), and intermediate state cells expressing both Foxi1 and Aqp4 (asterisks). The cell labeled 1 is the same cell in (E and G), and similarly the cell labeled 2 is the same cell in (E and G). (H) In the inner stripe of the OMCDs the average number of PCs per tubule length decreases significantly in Hes1cKO kidneys compared with wild-type controls. Intermediate cell types are present at significantly higher numbers in the Hes1cKO inner stripe of the OMCDs; *P<0.05, when performing unpaired t tests. A minimum of seven OMCD segments were analyzed per mouse, and three mice per genotypic group were analyzed. Scale bars, 10 μm.