|Home | About | Journals | Submit | Contact Us | Français|
The identity of niche signals necessary to maintain embryonic nephron progenitors is unclear. Here we provide evidence that Fgf20 and Fgf9, expressed in the niche, and Fgf9, secreted from the adjacent ureteric bud, are necessary and sufficient to maintain progenitor stemness. Reduction in the level of these redundant ligands in the mouse led to premature progenitor differentiation within the niche. Loss of FGF20 in humans, or of both ligands in mice, resulted in kidney agenesis. Sufficiency was shown in vitro where Fgf20 or Fgf9 (alone or together with Bmp7) maintained isolated metanephric mesenchyme or sorted nephron progenitors that remained competent to differentiate in response to Wnt signals after 5 or 2 days in culture, respectively. These findings identify a long sought-after critical component of the nephron stem cell niche and hold promise for long-term culture and utilization of these progenitors in vitro.
Different mammalian organs deploy different strategies to fulfill their physiological roles in the adult. The gut and the skin rely on stem cells sequestered in specialized niches to maintain homeostasis and recover from injury (Fuchs, 2008; Li and Clevers, 2010). In contrast, the adult kidney has no identifiable stem cells (Little and Bertram, 2009), displays limited repair capacity (Humphreys et al., 2008) and relies instead on a large surplus of nephrons generated from an embryonic pool of stem cells/progenitors during development. Genetic and environmental factors affect the number of nephrons; low numbers correlates with renal diseases and hypertension (Bertram et al., 2011; Keller et al., 2003; Myrie et al., 2011). Improved insight into the nephron progenitors and their environment may help maintain such cells for therapeutic purposes. Therefore, there is growing interest in understanding the mechanisms that regulate these cells during nephrogenesis.
Mammalian renal progenitors are established in the metanephric mesenchyme (MM) following ureteric bud (UB) outgrowth. The MM condenses around UB tips and induces tip branching (Costantini and Kopan, 2010); the distal aspect of the branched duct is a niche for embryonic multipotent stem/progenitor cells (Boyle et al., 2008; Kobayashi et al., 2008). Nephron progenitors located lateral or proximal to the UB tips form a pre-tubular aggregate (PTA) before undergoing mesenchymal to epithelial transition generating the renal vesicle (RV). The RV cells grow at different rates to form a S-shaped body (SSB). Nephron progenitors located distal to the UB form a “cap” and express the transcription factors Six2 and Sall1, which play a critical role in maintaining stemness. Six2-deficient progenitors differentiate en masse prematurely in a Wnt9b-dependent manner (Kobayashi et al., 2008; Self et al., 2006), causing renal agenesis. The UB is thought to play a dual role. It makes an important contribution to maintaining progenitors capable of self-renewal and it provides Wnt9b that promotes differentiation of cap metanephric mesenchyme (CMM) (Carroll et al., 2005; Karner et al., 2011; Kobayashi et al., 2008; Park et al., 2007). This stem cell/progenitor population disappears after producing 1.5×106 (human) or 1.3×104 (mouse) nephrons, likely due to the disappearance of their niche (Hartman et al., 2007)
Whether niche signals maintain stemness or block differentiation is unclear. Paradoxically, Wnt9b may constitute part of the niche signal (Karner et al., 2011). However, a niche “stemness” signal, if one exists, is expected to specifically maintain self-renewal while keeping the cells competent to respond to differentiation signals by generating multiple lineages. Such a molecule would have therapeutic potential if it could be used to maintain and expand a population of multipotent renal stem cells in an artificial niche.
FGF ligands are among the candidates for the niche signal required to maintain stemness because inactivation of FGF receptors (Fgfrs) in the CMM results in arrested kidney development (Poladia et al., 2006; Sims-Lucas et al., 2011). Supporting the notion that FGF signals are essential for maintenance of the nephron progenitor population is the observation that supplementing isolated MM cells with FGF2 (which is not present in the CMM) and Bmp7 (which is) can maintain their competence in culture short-term (Dudley et al., 1999). Generally, FGF ligands expressed in mesenchymal tissues signal to epithelial splice forms of FGFRs (‘b’ splice forms), whereas ligands expressed in epithelial cells signal to mesenchymal splice forms of FGFRs (‘c’ splice forms; (Zhang et al., 2006)). Based on this, the ligand may be provided to the CMM by an epithelial source, (ie the UB). CMM cells express Fgf7 and Fgf10, which signal to epithelial Fgfr2b in the UB. Targeted deletion of Fgf7 leads to a reduction in ureteric branch number due to decreased signaling through Fgfr2b and secondarily, to fewer nephrons (Bates, 2011; Qiao et al., 1999). Fgf8, expressed in the nascent RV, is required for progression from PTA to RV and for its proliferation (Grieshammer et al., 2005; Perantoni et al., 2005). Loss of Fgf8 results in apoptosis of MM due to loss of a PTA-derived signal, most likely Wnt4 (Grieshammer et al., 2005; Perantoni et al., 2005). Although these studies point to an important role for FGF signaling in maintaining the MM, the ligands that signal to mesenchymal Fgfrs in vivo have not been identified.
Here, we investigate the role of the Fgf9 sub-family in renal development and discovered that Fgf9 acts redundantly with Fgf20 to promote the proliferation, survival and stemness of CMM progenitors in vivo. Moreover, we identified a human missense mutation in FGF20 that results in bilateral renal agenesis. We show that either Fgf9 or Fgf20 were sufficient to maintain differentiation-competent, Six2+ nephron progenitors in vitro for 5 days, whereas cells grown in Fgf8 or Fgf10 lose competence by 48 hours. Sorted Six2+ cells retained competence to form epithelia for at least 48 hours in culture when grown in serum-free media containing Fgf9/BMP7. These studies identify Fgf20 (and Fgf9) as integral parts of the CMM niche where they act to maintain a self-renewing progenitor population competent to respond to an inductive cue. This work lays the foundation for creating an artificial renal stem cell niche.
The Fgf20 gene was targeted by insertion of a β-Galactosidase (βGal) cDNA into exon 1 (Huh et al., 2012). Fgf20βGal/βGal mice are viable, fertile and have a normal lifespan. They also display several phenotypes, including sensorineural hearing loss due to defects in the development of the organ of Corti (Huh et al., 2012). Examination of E18.5 kidneys from Fgf20βGal/βGal embryos revealed a reduction in kidney size (~15%) relative to control kidneys (n=8, P<0.03). Counts confirmed that fewer Glomeruli formed in Fgf20βGal/βGal and Fgf20+/βGal kidneys relative to wild type (15684±3035 vs. 22350±2733; p=8.1775×10-8; 19392±3070 vs. 22350±2733, respectively; P=0.0028; Figure 1I). The adrenal glands, gonads and Wolffian duct of Fgf20βGal/βGal were indistinguishable from wild type, as was the gross histology of Fgf20βGal/βGal kidneys (Figure 1B; see also Figure 2 and and3).3). The intermediate steps in nephron development (formation of the RV, SSB and early glomeruli) were comparable to control (white line, Figure 1F-G’, arrow SSB and arrowhead glomeruli). Thus, the overall smaller size was consistent with a smaller progenitor pool that differentiated normally.
Related FGFs have similar biochemical properties and developmental functions (Itoh and Ornitz, 2011; Zhang et al., 2006). We thus tested if the other members of the same subfamily (Fgf9, Fgf16) had redundant functions with Fgf20 in kidney development by generating compound Fgf9 and Fgf20 mutant animals. Fgf20 compound heterozygous (Fgf9+/−; Fgf20+/βGal and Fgf9−/−;Fgf20+/ßGal) embryos did not exhibit any gross abnormalities (n=10 and 6, respectively; Figure 1C) Fgf9-deficient embryos display aberrant development of the male reproductive tract with no kidney abnormalities (Colvin et al., 2001a), and die after birth due to lung malformations (Colvin et al., 2001b); therefore, all subsequent analyses were conducted at E18.5 unless stated otherwise. In contrast, all Fgf9+/−; Fgf20βGal/βGal embryos had hypoplastic kidneys (n=8). The defects ranged from mild (15% reduction in 8 kidneys, p= 0.005, Figure 1D’) to severe (60% reduction in 6 kidneys, P<0.001, Figure 1D”). One embryo displayed bilateral renal agenesis (similar to Figure 1E). No defects were detected in the adrenal glands, gonads, Mullerian ducts, ductus deferens or bladder. Significantly, histological analysis revealed that in Fgf9+/−; Fgf20βGal/βGal kidneys, large areas of the nephrogenic zone were missing, replaced by more mature tubular structures (Figure 1H-H’, arrowheads). Kidney agenesis was observed in all Fgf9−/−; Fgf20βGal/βGal embryos examined (n=6, Figure 1E, yellow arrowhead). In this genetic background, the adrenal glands were reduced in size (Figure 1E, white arrowhead).
These findings indicate that Fgf9 and Fgf20 function redundantly during kidney development to maintain the nephrogenic zone. Because one allele of Fgf20 in Fgf9 null embryos is enough to support normal kidney development, whereas Fgf9+/−; Fgf20βGal/βGal mice display a severe phenotype, Fgf20 seems to have a more dominant role than Fgf9 in the kidney.
To better characterize Fgf9 and Fgf20 deficient kidneys, we examined the expression of several molecular markers of nephron progenitors and their differentiated descendents in control, Fgf20βGal/βGal and Fgf9+/−; Fgf20βGal/βGal dysplastic kidneys. The UB (Cytokeratin 8, Ck8) branched and formed a collecting duct in both control and mutant mice (Figure 2A-E). Wilms tumor 1 (WT1) was highly expressed in the podocytes, and Lotus Tetragonolobus Lectin (LTL) positive proximal convoluted tubules (PT) were readily detected. This and the fact that some Fgf9+/−; Fgf20βGal/βGal mice survive to adulthood, confirmed the presence of mature, functional nephrons, albeit in reduced numbers within smaller, dysplastic kidneys (Figure 2A-E). Loss of Six2-positive CMM occurred in patches, consistent with an overall reduction in nephron progenitors (see below). As noted before, large areas in the nephrogenic zone of E18.5 Fgf9+/−; Fgf20βGal/βGal kidneys were replaced by mature structures (Figure 2C,E, white arrows). Six2-expressing nephron progenitors also express Pax2 (paired-box homeotic transcription factor 2) and low levels of WT1 protein. These progenitors form one or two cell layers distal to each UB tip in both control (Fgf20+/βGal) and in most tips in Fgf20βGal/βGal kidneys (Figure 2A-A’,D’ yellow arrowheads) at E18.5. In contrast, in Fgf9+/−; Fgf20βGal/βGal kidneys, three phenotypic categories can be discerned: normal CMM (Figure 2 yellow arrowheads), reduced progenitor numbers (Figure 2B’, C’, E’ white arrowheads, also seen near a few tips in Fgf20βGal/βGal kidneys), or loss of progenitors leading to mature structures distal to UB tips (white arrows in Figure 2C-C’,E’).
To determine if these defects were intrinsic to the progenitors, we cultured E11.5 metanephroi of various genotypes for 2 days and stained for Ck8 and Six2. Fgf20βGal/βGal metanephroi were smaller than controls. To evaluate progenitor pool size, we counted the number of tips surrounded by Six2+ cells in each genotype (see experimental procedures for detail). Whereas UB tips in control animals were surrounded with Six2+ cells, 33.7% of Fgf20βGal/βGal UB tips were surrounded by fewer Six2+ cells (yellow arrows, Figure 3C) and 12.4% lacked Six2+ cells altogether (blue arrows, Figure 3c; quantification in M). Fgf9+/−; Fgf20βGal/βGal metanephroi displayed a more severe phenotype (41% had fewer and 29.8% had no Six2+ cells; Figure 3D-F, M). Staining for the RV-SSB differentiation marker Jagged1 (Jag1) (Cheng et al., 2003; Cheng et al., 2007) detected numerous Jag1+ epithelial differentiating distal to the UB tip, where Six2-expressing mesenchyme should reside (Figure 3H-L, white arrows). Consistent with premature differentiation in the niche, Wnt4 was ectopically expressed in cells surrounding UB tips in Fgf9+/−; Fgf20βGal/βGal metanephroi (Figure 3K, white arrows). These observations are consistent with the hypothesis that Fgf9 and Fgf20 act to prevent premature differentiation of nephron progenitors, reminiscent of Six2-deficient mice (Self et al., 2006) or mice expressing constitutively active Notch in the CMM (Boyle et al., 2011; Cheng et al., 2007; Fujimura et al., 2010).
To determine the temporal and spatial expression patterns of Fgf20, we analyzed FGF20-βGal activity during kidney development. Whole-mount X-gal staining at E10.5-12.5 clearly showed activity in mesonephric and metanephric kidneys (Figure 4A-C, black and red arrowheads, respectively). At E14.5, βGal activity was present in the MM surrounding the UB (Figure 4D). To identify the specific cell population that expressed Fgf20, we stained sections of Fgf20βGal/βGal kidneys with antibodies against Ck8, Neural Cell Adhesion Molecule (NCAM, marking mesenchymal cells and their epithelial derivatives) and βGal. At E16.5, βGal co-localized with Six2 in CMM cells (Figure 4E-F, white arrows). βGal activity persisted in the nascent RV due to expression or perdurance (Figure 4E, yellow arrowhead) but not in the stroma or the UB (Figure 4E-F). We therefore conclude that in the metanephric kidney, Fgf20 is expressed exclusively in nephron progenitors. This pattern agrees with the mRNA expression data collected by GUDMAP (Little et al., 2007) and presented in (Brown et al., 2011). The same resource describes Fgf9 expression as most abundant in the UB, with MM expression detected P0-P2 (see Figure S1). mRNA for both ligands disappears as the remaining progenitors differentiate en-masse.
Conditional inactivation of Fgfrs in the CMM caused renal agenesis (Poladia et al., 2006), mimicked by Fgf9/20 compound mutants (Figure 1). Because Fgf9 is expressed in the duct (Barasch et al., 1997; Brown et al., 2011; McMahon et al., 2008), we hypothesized that Fgf9 is a paracrine signal from the duct to the CMM and FGF20 acts in an autocrine manner on the CMM itself. To test this, we inactivated a conditional allele of Fgf9 (Fgf9f/f; (Lin et al., 2006)) specifically in the UB, with the HoxB7-Cre BAC transgene (Yu et al., 2002). Removal of both Fgf9 alleles from the UB in Fgf20+/βGal embryos permitted normal development (Figure 4G-H,K-K’). Conversely, when a single allele of Fgf9 was removed from the UB of Fgf20βGal/βGal embryos (HoxB7-Cre+/tg; Fgf9+/f; Fgf20βGal/βGal), dysplastic, smaller kidneys formed (25-75%; n=10; Figure 4I). Examination of these kidneys (Figure 4L-L’) revealed depletion of nephron progenitors similar to that seen with global loss of an Fgf9 in this background (Figure 2). Fgf20 null embryos lacking both Fgf9 alleles in the UB (HoxB7-cre+/tg; Fgf9f/f; Fgf20βGal/βGal) displayed rudimentary kidneys (n=6) or renal agenesis (n=4) (Figure 4J), a phenotype slightly milder than the Fgf9/Fgf20 compound null. To ask if Fgf9 was expressed in the MM, we deleted Fgf9 from the mesenchyme using the Pax3-Cre allele (Cheng et al., 2007; Li et al., 2000). Some Pax3-Cre+/tg; Fgf9f/f; Fgf20βGal/βGal mice have rudimentary kidneys (Figure S1), consistent with Fgf9 being expressed in the mesenchyme. These data support a model in which Fgf9 secreted from the UB complements the activity of Fgf20 and Fgf9 produced in the CMM.
Independently, we used a combined approach of homozygosity mapping and exome sequencing to analyze a consanguineous family with several fetuses presenting with isolated bilateral renal agenesis. Autopsy revealed the Potter sequence and bilateral reno-ureteral agenesis (Figure 5A) in one fetus. Other organs were normal. DNA analysis from this fetus identified homozygous variants in four genes expressed during early kidney development and located in homozygous chromosomal regions. One of these mutations was a single base-pair deletion in exon 2 of FGF20, resulting in a frameshift starting at amino acid 113 and terminating in a stop codon at position 121, before the third Heparin-binding domain of the protein (Figure 5B-D, mutations (c.337delG)). Sanger sequencing analysis confirmed homozygosity for the FGF20 c.337delG mutation in this fetus. Moreover, we showed that the mutation was homozygous in two other affected fetuses in the family and heterozygous in the parents with available DNA (Figure 5B). This mutation was absent in a healthy sibling (Figure 5B), from dbSNP and the 1000 genome database, and was not found in our in-house whole exome sequence data. Moreover, the three other homozygous variants identified in this fetus did not segregate with the kidney defect in this family. Combined with the mouse mutants described above, these data point to FGF20 as an essential protein during mammalian metanephric kidney development whose loss is sufficient to cause bilateral renal agenesis in humans.
Renal epithelia fail to form in humans lacking FGF20 or in mice lacking Fgf9 and Fgf20. We next asked if kidney agenesis reflected failure to establish the CMM (Costantini and Kopan, 2010) or failure to maintain the progenitors. Since the CMM induces UB outgrowth, we asked if UB induction was initiated properly in mutants by whole-mount Ck8 staining. By E11.5, control embryos had an elongated, T-shaped UB (Figure 6A). Fgf9−/−; Fgf20βGal/βGal embryos initiated UB outgrowth, but in contrast, we only observed a branched UB in 2 out of 7 embryos (Figure 6B). To examine the MM, transverse sections of E11.5 embryos were prepared. In all embryos, the elongating UB contacted the Pax2-positve MM (Figure 6C-H). Control MM cells condensed around the UB in wild type embryos (Figure 6C), whereas the Fgf9−/−; Fgf20βGal/βGal MM was significantly smaller and most cells failed to coalesce around the UB (Figure 6D). TUNEL staining detected many apoptotic cells within the MM of Fgf9−/−; Fgf20βGal/βGal embryos at E11.5 (Figure 6E-F). By E12.5, only a few Pax2+, Ck8− cells remained in the mesenchyme (Figure 6G-H). This analysis suggested that, in the absence of Fgf9 and Fgf20, UB outgrowth was initiated, but the rudimentary MM failed to migrate towards the UB and, in addition to prematurely differentiating (Figure 3 L-K), underwent high rates of apoptosis reminiscent of isolated MM cells in vitro and of Fgfr-deficient MM in vivo (Poladia et al., 2006). Analysis of Pax2 and Gdnf expression in the MM correlated to the dose of Fgf9/20 as early as E10.5-11.5. In contrast, Pax2 (Figure 6) and cRet expression (data not shown) in the UB was unchanged. It cannot be ruled out that Fgf9 has a function in maintaining Etv4 and Etv5 in the UB epithelium (data not shown). The consequence of reduction in GDNF/ret signaling was progressive failure in branching morphogenesis (data not shown), which could be reversed in vitro by the addition of GDNF to metanephroi with limited branching (Figure S2).
The data presented thus far shows that Fgf20 and Fgf9 together are required for maintenance and survival of nephron progenitors in mice, and infer a similar activity for FGF20 in humans. Moreover, our allelic series indicates that stemness and survival were sensitive to the dose of Fgf9/20 (predicting reduced nephron numbers in humans heterozygous for FGF20). To test if these ligands were sufficient to mimic the niche, we examined the ability of recombinant FGF9 to preserve the stemness of isolated nephron progenitors in vitro. MM was isolated from E11.5 kidneys and explants were grown in serum free media supplemented with Heparin (an important cofactor for FGF signaling; (Yayon et al., 1991)), or Heparin plus FGF9. In this and subsequent experiments, the absence of UB was confirmed by Ck8 staining. To assess survival, we used TUNEL staining; to identify mitotic cells, we used anti phospho-histone H3 (pHH3) antibody. As reported, the majority of isolated MM cells were TUNEL-positive after 4 days in media (Figure 7A). When Heparin was added, a greater number of cells survived and some mitotic cells were seen (Figure 7B). The addition of FGF9 and Heparin improved both survival and proliferation (Figure 7C). Anti-Cited1, Wt1 (not shown) and Six2 antibody staining 48hr after MM isolation confirmed that these markers were maintained in explants cultured in Heparin/FGF9-containing media (Figures 7E-E’), but not with Heparin alone. Some Six2-expressing cells were mitotic (Six2+; pHH3+; Figure 7F-I). These data demonstrate that the addition of FGF9 was sufficient to promote survival and support proliferation of some nephron progenitors in-vitro.
Next we asked whether nephron progenitors grown in the presence of FGF9 retained their competence to respond to inductive signals. To avoid introduction of renal tissue, we used freshly dissected dorsal spinal cord segments (dSC; (Saxen and Lehtonen, 1987; Saxen and Sariola, 1987)) as a source of Wnt (Herzlinger et al., 1994; Kispert et al., 1998). After 48hrs in Heparin/FGF9 media, dSC was added and the cells were maintained for an additional 2-4 days in serum free media. All freshly isolated control MM grown in contact with the dSC underwent MET (data not shown). After two days in vitro, only FGF9-supplemented MM differentiated into epithelia expressing either early distal (Cdh1) or proximal (Jag1) markers and retaining a few Six2+ progenitors (Figure 7J-K). The fact that these tubular structures contained at least two differentiated cell types (distal and proximal epithelial cells) strongly suggests that progenitor multipotency was retained.
It is clear that Fgf20 plays a more dominant role than Fgf9 in vivo and that any other Fgfs present in the cortex during development (Fgf2, Fgf8 and Fgf10) cannot rescue the niche in the absence of Fgf9 and Fgf20. Fgf9 and Fgf20 share biochemical properties that may be unique among Fgf ligands expressed in the vicinity of the niche. To test this possibility, we cultured isolated E11.5 MM for 48hrs in Heparin supplemented media and added FGF9, FGF20, FGF8 or FGF10. First, we established the activity of our recombinant proteins on FGFR2b and FGFR2c splice variants stably expressed in Baf3 cells (Ornitz et al., 1996) Very few Six2-expressing cells (~10 cells/explants) were seen in the presence of FGF8 (7/7 explants); none were detected in eight explants exposed to FGF10 (Figure S3). In summary, these data demonstrated that members of the Fgf9/20 sub-family are necessary and sufficient to promote survival, competence and multipotency of MM cells in vitro.
FGF2 and BMP7 synergized to improve maintenance of nephron progenitors in vitro (Dudley et al., 1999). To test if BMP7 can enhance the effect of the endogenous niche ligand, we cultured isolated MM in media supplemented with BMP7 and FGF9. Whereas isolated MM supplemented with BMP7 alone lost their Six2+ population (Figure 8A), BMP7/FGF9 seemed to increase the overall number of Six2+ cells and induced nephron progenitors to sort away from Six2− cells and aggregate, confounding efforts to count them (Figure 8B-B’). In the presence of dSC, numerous epithelial structures containing distal (Cdh1+) and proximal (Jag1+) cells formed, each surrounded by a tight cluster of Six2+ cells (Figure 8C-D). To assess for self-renewal of nephron progenitors in media supplemented with BMP7 and FGF9, EdU (5-ethynyl-2’-deoxyuridine) was added to freshly isolated MM for 4-7hr, washed and refed the cells every 12hrs. After two days many Six2+ cells that remained EdUbright (Figure 8E-H, yellow arrows) were detected, suggesting that they only divided once. A few Six2+, EdUDimm cells (progenitors that divided once after labeling or labeled late in S phase) were in mitosis after 48hrs (stained by pHH3, Figure 8E-H, turquoise arrow). Since mammalian S, G2 and M phases are relatively fixed in length, such double label indicates that some nephron progenitors were able to divide twice (one immediately after culture, the other two days later) in the presence of BMP7 and FGF9.
To ask if Six2+ cells could be maintained for as long as they live in vivo (from E11.5 to P2, or 8 to 10 days), we cultured isolated MM with FGF9 and BGF7 for 8 days in culture. Notably, although some Six2 expressing cells were in mitosis (Six2+; pHH3+; Figure S4 A-C, yellow arrows), the proportion of Six2+ cells declined after 5 days in these conditions (Figure 8 I, I’); by day 8, only a few Six2+ cells were detectable (data not shown). Staining for differentiation markers after 5 days confirmed that no Cdh1 or Cdh6 expressing epithelia formed in the absence of inducers (data not shown). Importantly, after 5 days in Heparin/FGF9/BMP7 media in vitro, Six2+ nephron progenitors formed epithelial tubules when co-cultured with dSC for an additional 4 days in basal media (9/9, three individual MM each in three repeats). These epithelia expressed either proximal (Cdh6) or distal (Cdh1) markers (Figure 8J).
Finally, to determine if FGF9 acted directly on Six2+ cells or indirectly through FoxD1+ stromal cells, we sorted GFP+ cells by FACS from Six2-GFPCREtg/+ kidneys. Small colonies of Six2+ cells, maintained for 48hrs in serum free media supplemented with FGF9, remained competent to differentiate (Figure 7L). Culturing these cells with FGF9+BMP7 for 48hr resulted in robust expansion of many more Six2+ colonies (Figure 8K-K’) that maintained competence to respond to differentiation signals and produced epithelia expressing either proximal (Cdh6) or distal (Cdh1) markers after co-culture with dSC (Figure 8L). Mitotically active nephron progenitors (Six2+; pHH3+) survived after 5 days (Figure S4E-F, yellow arrows). Although when cultured with dSC they could form a few Cdh1 or Cdh6 positive cells, tubular epithelia no longer formed (data not shown). These data demonstrate that Fgf9/Bmp7 act directly on progenitors, promoting their survival while maintaining their competence to respond to inductive signals for at least 2 days.
Here we present data demonstrating that members of the Fgf9 ligand family are bona fide secreted niche signals in the developing mammalian kidney. FGF20 acts alone in humans but is redundant with Fgf9 in mouse. In the absence of Fgf9 and Fgf20, UB outgrowth initiates, but because the MM undergoes a high rate of apoptosis, or because Fgf9 signals needed to maintain Etv4 and Etv5, UB branching is halted and kidney development fails. In human fetuses homozygous for a truncated allele of FGF20, no kidneys form. In mice, a single allele of Fgf20 (Fgf9−/−; Fgf20+/βGal) is sufficient to maintain normal kidney development in vivo (however, not in isolated MM), whereas a single allele of Fgf9 (Fgf9+/−; Fgf20βGal/βGal) is insufficient. The redundancy between Fgf9 and Fgf20 produced highly informative variable phenotypes ranging from near normal kidneys to renal agenesis. When Fgf9 and Fgf20 are at limiting doses, the balance between differentiation and self-renewal is disturbed, Six2 expression is reduced, and precocious CMM differentiation can occur suggesting a role for FGF signaling in the maintenance of nuclear factors necessary for stemness (Osafune et al., 2006; Sakaki-Yumoto et al., 2006; Self et al., 2006). Consequently, many Fgf9+/−; Fgf20βGal/βGal kidneys contain regions where nephron progenitors were replaced by differentiating epithelia, reminiscent of MM lacking Six2 (Kobayashi et al., 2008; Self et al., 2006). Members of the Fgf9 sub-family are unique in that they can signal to both ‘b’ and ‘c’ splice forms of some Fgfr (Figure S3E and (Ornitz et al., 1996; Zhang et al., 2006)). However, autocrine Fgf9 also regulates epithelial growth and branching (del Moral et al., 2006; White et al., 2006; Yin et al., 2011; Yin et al., 2008). Although genetic deletion confirmed that mesenchymal Fgf9 makes a measurable contribution to the niche, the bulk of Fgf9 signals are provided from the epithelia (UB) to mesenchyme (MM). In contrast, Fgf20 was expressed exclusively in the CMM, signaling in an unusual autocrine fashion to maintain the mesenchymal cells in which it is expressed. Interestingly, in the developing organ of Corti, Fgf20 provides an autocrine signal acting within the epithelium to regulate its differentiation (Huh et al., 2012).
Independently, other investigators observed that Bmp7, expressed in nephron progenitors, UB and PTA during early kidney development (Dudley et al., 1995), acts to promote survival and proliferation of nephron progenitors (Blank et al., 2009; Dudley et al., 1999; Dudley et al., 1995; Dudley and Robertson, 1997; Godin et al., 1998). In Bmp7 null mice, the first round of nephrogenesis is induced, followed later by loss of nephron progenitors, resulting in premature termination of kidney development (Dudley et al., 1995). Moreover, Fgf2, which like Fgf9 is expressed in the UB, can promote survival and maintain competent MM cells in vitro (Barasch et al., 1997; Dudley et al., 1999; Perantoni et al., 1995). However, FGF2 is not necessary in the niche as Fgf2 null mice do not display a kidney phenotype (Zhou et al., 1998). FGF2, FGF8 and FGF10 could not rescue Fgf9/20 null embryos, indicating that their function does not overlap with Fgf9/20 during kidney development. FGF7 and FGF10 signal to the UB and can replace GDNF when the inhibitor Sprouty1 is removed (Michos et al., 2010). Fgf8 is expressed in the RV and in renal epithelia and is needed for their maintenance (Grieshammer et al., 2005; Perantoni et al., 2005). In vitro analysis confirmed that FGF8 and FGF10 cannot maintain isolated MM, perhaps due to higher affinity of Fgf8 to the decoy receptor Fgfrl1, which is expressed in CMM (Brown et al., 2011). Moreover, several other ligands capable of activating the PI3K pathway can act as well as Fgf9 in maintenance of the progenitors (Brown et al., 2011), but our in vivo analysis demonstrates that none can rescue the loss of Fgf20 and Fgf9.
In contrast to Fgf8, either Fgf9 or Fgf20 are sufficient to support a Six2+ cell population that remained competent to respond to dSC by differentiating into different cell types. FGF9 maintains sorted, purified progenitors for at least 5 days; this is only possible if Fgf9 acted directly on the progenitors and rules out an indirect role via an essential, Fg9-dependent UB or stromal factor. Although restoring the unique cohesion of CMM cells in vitro also requires Bmp7, this cohesion is not needed for retention of competence; Fgf9 alone was sufficient to maintain Six2+ progenitors that are competent to differentiate in response to Wnt. Nonetheless, Bmp7 helps to organize cells into a structure that resembles the CMM niche, perhaps maximizing the benefits of endogenous Fgf20. At present, it is unclear if conditions that allow robust growth rates are compatible with maintaining competence as highly proliferative Six2+ cells in nephrospheres lost their competence to differentiate (Lusis et al., 2010).
It is thought that Wnt9b signals maintain stemness where Six2 is present and drives differentiation where Six2 is absent (Carroll et al., 2005; Karner et al., 2011; Kobayashi et al., 2008; Self et al., 2006). Although it is clear that self-renewing nephron progenitors receive a Wnt9b signal within the CMM (Karner et al., 2011), it is also evident that in the absence of Wnt9b, nephron progenitors survive; they do not undergo apoptosis, and they remain competent to respond to inductive signals as late as E13.5 (Karner et al., 2011)). These observations indicate that signals other than Wnt9b maintain competent nephron progenitors in Wnt9b mutant mice, and that in the absence of Wnt9b, proliferation is attenuated and differentiation is lost. Fgf9 and Fgf20 could provide this signal; both are Wnt-responsive genes in some contexts (Chamorro et al., 2005; Karner et al., 2011). The details of the relationships between Wnts and FGFs within the MM remain to be elucidated.
In conclusion, these observations suggest that, at a minimum, Fgf9/20 and Bmp7 organize the nephron progenitor niche and highlight the essential role of FGF20 in human kidney development. Regulating the balance between self-renewal and differentiation is key in determining nephron number, and lower nephron numbers are correlated with higher risk for hypertension, kidney disease, and recovery from injury. Our data indicate that FGF signaling likely regulates multiple important steps in the niche, including survival, proliferation, and competence. The relative contribution of each of these to maintenance of a proper niche is unclear. Since nephron progenitors do not persist in the adult, understanding the mechanisms that regulate this niche is a key step in the development of new therapeutic cell-replacement approaches. The identification of the endogenous niche ligands (Fgf9 and Fgf20) and the demonstration that they can maintain purified progenitors for 5 days in vitro opens the possibility for long-term maintenance of self-renewing nephron progenitors in culture. The importance of such a culture for research is self-evident; the therapeutic potential of these cells will have to be empirically determined in various injury models.
All mice were maintained in the Washington University animal facility according to animal care regulations; Animals Studies Committee of Washington University approved the experimental protocols (protocol number 20110027). Lines used here are: Fgf20βGal (Huh et al., 2012); Fgf9− (Colvin et al., 2001a); Fgf9f (Lin et al., 2006); HoxB7-Cre+/tg (Yu et al., 2002) and Six2-GFPCREtg/+ (Park et al., 2007). Mice and embryos were genotyped using the universal PCR genotyping protocol (Stratman et al., 2003). Primers sequences available upon request.
Isolation and culture of metanephric kidneys were performed as described (Barak and Boyle, 2011). To manually isolate the MM, E11.5 metanephric kidneys were dissected in cold PBS. Kidneys were treated with 2.25% pancreatin and 0.75% trypsin in Ca-Mg free Tyrode’s solution (pH7.4) (P/T solution) for 30s, then washed with FBS. MM was dissected from the UB using 26G needles and placed on transwell filters. Explants were cultured at 37°C, 5% CO2 in defined medium (Barak and Boyle, 2011) supplemented with the following as indicated. 1μg/ml Heparin (Sigma), 8.6nM FGF9 (PeproTech), 50ng/ml BMP7 (R&D systems) and 100ng/ml FGF2 (PeproTech for positive control, data not shown). FBS was added as carrier in stock solutions (same results were obtained with BSA as carrier). Media was changed daily and activity of FGFs protein was tested by a Baf3 assay as described (Ornitz et al., 1996). To examine competence to respond to inductive signal, dSC was isolated from E11.5-13.5 embryos. Two fresh dSC segments were placed adjacent to each MM explant. Explants were then cultured for an additional 2-4 days in unsupplemented media. At least 3 explants per treatment were prepared for each experiment. Experiments were repeated at least 3 times on different days. Staining for Ck8 was performed to confirm exclusion of UB remnants from explants.
Kidneys were dissected from E14.5-E17.5 embryos from wild type females mated with Six2-GFPCREtg/+ males. Both GFP positive and negative kidneys were collected, with negative kidneys serving as sorting controls. To dissociate the cells into single cell suspension, kidneys were treated with pancreatin/trypsin diluted in Tyrode s solution (Ca+2, Mg+2 free) for 15min on ice (200ul per 4-8 kidneys), followed by pipetting until clumps were not visible. To stop digestion, 1ml of cold DMEM with 10% FBS was added to the 200ul of enzyme solution. Cell suspension was then filtered through a 50μm filter (CellTrics®-Partec disposable filters 04-004-2327), collected in 1.5ml tubes and centrifuged for 4min at 800rpm. Media was aspirated and cells were resuspended in FACS buffer (3% FCS in PBS). GFP+ cells were sorted using a MoFlo high-speed flow cytometer (Dako Cytomation, Fort Collins, CO). Gating was implemented based on negative control profiles to select for live GFP positive cells. Sorted cells were collected in 1.5ml tubes (50,000-100,000 cells per tube) containing kidney media with 5% FBS and kept on ice before plating. Finally, cells were spun down for 4min at 800rpm and plated directly on transwell filters in organ culture dishes with kidney media supplemented with Heparin containing Fgf9 or Fgf9 and Bmp7. Media was changed every 12hr.
Immunohistochemistry was performed as detailed in supplemental data.
Two-tailed unpaired t-tests were performed for statistical analysis. Resulting p values are noted in text. For comparaison of kidney sizes, perimeter was measured and used for statistical analysis. Images were acquired with a Zeiss Axioimage Z1 equipped with an ApoTome or with stereomicroscope Leica MZ10F.
Kidneys were dissected from P5 pups of Fgf20 WT (+/+), Fgf20 +/βgal (+/-) and Fgf20 βgal/βgal (-/-) mice. To dissociate the tissue, kidneys were chopped into 2mm2 pieces and digested in 5ml of 6N HCL at 37°C for 90min. Tissue was further dissociated by repeated pipetting. 25ml of ddH2O was added to each sample, and samples were kept at 4°C. To count the number of nephrons, 100μl of well-mixed suspension was placed onto a 1cm×1cm area in a p100 culture dish marked with gridlines. The number of glomeruli were counted under an inverted microscope 5 times. Total number of glomeruli was estimated as follows: Total nephron number per pair of kidneys= average number of glomeruli in 100ul × 300.
The fetus studied by exome sequencing (fetus 2 in Figure 5B) was an 18-week gestation male terminated for recurrence of bilateral renal agenesis with anhydramnios. The couple was consanguineous (first cousins) Caucasian; the mother was a healthy 25-year-old woman with previous history of termination of pregnancy at 26 weeks for anhydramnios. In this family, another consanguineous couple also experienced two terminations of pregnancy for bilateral renal agenesis (fetuses 3 and 4 in Figure 5B). At autopsy, fetus 2 was eutrophic, presented the Potter sequence including redundant skin, varus feet and pulmonary hypoplasia. Internal examination showed bilateral reno-ureteral agenesis (Figure 5A). Other organs were normal. X-ray evaluation showed no anomaly for the gestational age. Neuropathological examination did not detect any abnormality in brain or eyes. DNA was extracted from a frozen liver sample of fetus 2. Genotyping was performed on Illumina Infinium HumanOmni1 beadchip, resulting in the identification of 26 homozygous genomic regions extending from 35 Mb to 700 Kb and spanning the genome (data not shown). Whole exome sequencing was performed using the 50 Mb Agilent SureSelect assay and a SOLiD4 sequencer (50 base fragment reads). The 1.4 Gb of mappable sequence data were aligned to the human genome reference sequence (hg19 build) using BWA aligner (Li and Durbin, 2010). Variants were annotated with in house pipeline based on the Ensembl database (release 61). Known variants from dbSNP132, the 1000 Genomes Project and in-house exome data were excluded from the genetic variants identified in the coding regions or splice sites. Subsequent filtration of the data was performed in order to select variants located in the identified homozygous genomic regions, in genes expressed in the early stages of kidney development (http://www.gudmap.org/) and with a predicted deleterious effect on the protein (Polyphen, Sift). Mutations identified by exome sequencing were validated by Sanger sequencing in all available DNA samples of the family: fetus 2, his parents, a healthy sister and fetuses 3 and 4. The study was approved by the Comité de Protection des Personnes pour la Recherche Biomédicale Ile de France 2 and informed consent was obtained from the parents.
We thank the family with renal agenesis for their participation in the study. We wish to thank Drs. Fen Wang (Fgf9f/f mice), Andrew McMahon (Six2 antibody, Six2-GFPCre and HoxB7-Cre mice), Frank Costantini, Masato Hoshi, Sanjay Jain and Gail Martin for sharing reagents and transgenic animals; Scott Boyle, Kamesh Surenduran, Chihiro Sato and Courtney Karner for discussions and critical reading of the manuscript; Ling Li for assistance with FGF activity assays; Tao Shen and Hannah Weis for assistance with animal husbandry and mouse genotyping; Mary Fulbright for expert technical help; and the Department of Developmental Biology Histology Core for preparation of tissue sections. We thank William Eades, Christopher Holley and Jacqueline Hughes at the Siteman Cancer Center High Speed Sorter Core Facility for performing cell-sorting; Jeanne Amiel for providing DNA from two fetuses; Wassila Carpentier for genotyping analysis; and Solenn Pruvost and Mohamed Zarhrate for technical assistance. RK, SC and HB were supported by a grant from the National Institutes of Diabetes and Digestive and Kidney Disease (DK066408) and the O’Brien Center Grant (5P30DK079333 to HB) provided support for organ culture experiments. The Siteman Cancer Center is supported in part by NCI Cancer Center Support Grant P30 CA91842. This work was supported in part also by a grant from the March of Dimes (DMO) and NIH support grants, P30DC04665, P30DK052574 and P30AR057235, for production of mice. Human sequencing analysis was supported by the GIS-Institut des Maladies Rares program (AAE11010KSA to CJ).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.